feat(fast): /fast ultrafast for GPT-6 Astra, priced at the served tier

OpenAI's Ultrafast service tier (service_tier="ultrafast") is broadly
available for GPT-6 Astra in the API and Codex (Pro 500 / Enterprise),
billed at 6x Standard. GPT-6.1 Sol Ultrafast is announced as coming soon.

- One tier table: agent.fast_mode.parse_service_tier / STATIC_TIERS now
  back the CLI, gateway and TUI config loaders (three hand-copied parsers).
- resolve_fast_mode_overrides(tier="ultrafast") sends the tier only for
  Ultrafast models on api.openai.com / chatgpt.com; anything else gets
  nothing rather than a silent swap to another paid tier.
- /fast ultrafast on CLI, gateway (typed + picker when supported) and TUI;
  static-tier attach sites pass the tier through.
- Pricing: the response's SERVED service_tier is folded into usage (Codex
  stream assembler keeps it); served ultrafast prices from a published
  6x row (272K whole-request tier included). A request asking for
  Ultrafast but served at default bills at Standard.
- i18n: 4 new keys in all 17 locales; option lists mention ultrafast.

agent/aux_accounting.py

@@ -64,9 +64,9 @@ def record_aux_usage(
         if raw_usage is None:
             return
 
-        from agent.usage_pricing import estimate_usage_cost, normalize_usage
+        from agent.usage_pricing import estimate_usage_cost, normalize_usage, with_served_service_tier
 
-        usage = normalize_usage(raw_usage, provider=provider)
+        usage = with_served_service_tier(normalize_usage(raw_usage, provider=provider), response)
         if not (
             usage.input_tokens or usage.output_tokens
             or usage.cache_read_tokens or usage.cache_write_tokens

agent/codex_runtime.py

@@ -779,7 +779,7 @@ def _output_text_of(item: Any) -> str:
 class _CodexResponseAssembler:
     """Assemble a Response-shaped ``SimpleNamespace`` from raw Responses SSE events.
 
-    Only ``usage`` / ``status`` / ``id`` are read from the terminal frame — never ``response.output``. Output
+    Only ``usage`` / ``status`` / ``id`` / ``service_tier`` are read from the terminal frame — never ``response.output``. Output
     items come from ``output_item.done``, or are synthesized from text deltas, or settled from function calls
     announced via ``output_item.added`` but never confirmed (some backends omit per-item done events on success)."""
 
@@ -790,6 +790,7 @@ class _CodexResponseAssembler:
     active_summary_index: Any = None
     terminal_status: str = "completed"
     terminal_usage = terminal_response_id = terminal_incomplete_details = terminal_error = None
+    terminal_service_tier = None  # the tier the backend SERVED (may differ from the one requested)
     # terminal_status defaults to "completed", so settlement needs an explicitly observed response.completed frame.
     saw_response_completed = False
 
@@ -911,6 +912,7 @@ class _CodexResponseAssembler:
         resp_obj = _event_field(event, "response")
         if resp_obj is not None:
             self.terminal_usage, self.terminal_response_id = _event_field(resp_obj, "usage"), _event_field(resp_obj, "id")
+            self.terminal_service_tier = _event_field(resp_obj, "service_tier")
             rstatus = _event_field(resp_obj, "status")
             if isinstance(rstatus, str):
                 self.terminal_status = rstatus
@@ -978,7 +980,7 @@ class _CodexResponseAssembler:
         return SimpleNamespace(
             output=output, output_text="".join(self.text_deltas), usage=self.terminal_usage, status=self.terminal_status,
             id=self.terminal_response_id, model=self.model, incomplete_details=self.terminal_incomplete_details,
-            error=self.terminal_error)
+            error=self.terminal_error, service_tier=self.terminal_service_tier)
 
 
 def _consume_codex_event_stream(

agent/fast_mode.py

@@ -1,7 +1,7 @@
 """Bounded fast-mode windows (``/fast auto`` and ``/fast cold``).
 
-``agent.service_tier``: ``None`` (normal), ``"priority"`` (static fast, pinned into
-``agent.request_overrides`` at build time), ``"auto"`` (every user turn opens a
+``agent.service_tier``: ``None`` (normal), ``"priority"`` / ``"ultrafast"`` (static tiers,
+pinned into ``agent.request_overrides`` at build time), ``"auto"`` (every user turn opens a
 window of ``agent.fast_auto_seconds``) or ``"cold"`` (only a session's first turn,
 no prior history, opens it). The provider's fast override is layered onto request
 kwargs only while the window is open; only per-request params (``service_tier`` /
@@ -20,6 +20,27 @@ DEFAULT_WINDOW_SECONDS = 60
 # Documented fast-mode rate-limit headers; a limit of 0 means the organization has no fast
 # capacity for the model (https://platform.claude.com/docs/en/build-with-claude/fast-mode).
 _FAST_LIMIT_HEADERS = ("anthropic-fast-input-tokens-limit", "anthropic-fast-output-tokens-limit")
+#: Tiers sent on every request of the session (OpenAI ``service_tier`` values; ``priority`` also
+#: selects Anthropic/xAI fast mode). Ultrafast is OpenAI-only and gated per model.
+STATIC_TIERS = frozenset({"priority", "ultrafast"})
+NORMAL_TIER_WORDS = frozenset({"", "normal", "default", "standard", "off", "none"})
+# User/config word -> agent.service_tier. The single table every surface (config loaders, /fast
+# on CLI / gateway / TUI) parses through, so a new tier is one edit.
+SERVICE_TIER_WORDS: dict[str, str] = {
+    "fast": "priority", "priority": "priority", "on": "priority",
+    "ultrafast": "ultrafast", "auto": "auto", "cold": "cold",
+}
+
+
+def parse_service_tier(raw: Any) -> str | None:
+    """``agent.service_tier`` for a user/config word; None for normal and for unknown words."""
+    value = str(raw or "").strip().lower()
+    return None if value in NORMAL_TIER_WORDS else SERVICE_TIER_WORDS.get(value)
+
+
+def service_tier_word(tier: Any) -> str:
+    """The user-facing word for a stored tier (``priority`` -> ``fast``, None/"" -> ``normal``)."""
+    return {"priority": "fast", None: "normal", "": "normal"}.get(tier, tier)
 
 
 def begin_turn(agent: Any, conversation_history: Any) -> None:

agent/turn_usage.py

@@ -17,7 +17,7 @@ from typing import Any, Dict, List
 
 from agent.image_token_cost import calibrate_from_usage
 from agent.usage_anchor import capture_usage_anchor, set_usage_anchor
-from agent.usage_pricing import estimate_usage_cost, normalize_usage
+from agent.usage_pricing import estimate_usage_cost, normalize_usage, with_served_service_tier
 
 logger = logging.getLogger("agent.conversation_loop")
 
@@ -98,7 +98,8 @@ def record_response_usage(
         )
         return ResponseUsageOutcome(compression_attempts=compression_attempts, rearmed=rearmed)
 
-    canonical_usage = normalize_usage(response.usage, provider=agent.provider, api_mode=agent.api_mode)
+    canonical_usage = with_served_service_tier(
+        normalize_usage(response.usage, provider=agent.provider, api_mode=agent.api_mode), response)
     # Aggregator-only usage kept for pricing: advisor tokens are priced at each advisor's
     # OWN model rate and added as dollars below.
     aggregator_usage = canonical_usage

agent/usage_pricing.py

@@ -2,7 +2,7 @@ from __future__ import annotations
 
 import logging
 import re
-from dataclasses import dataclass, fields
+from dataclasses import dataclass, fields, replace
 from datetime import datetime, timezone
 from decimal import Decimal
 from typing import Any, Dict, Literal, Optional
@@ -290,6 +290,22 @@ for _slug, _inp, _out, _read, _write, _inp_above, _out_above, _read_above, _writ
     )
 del _slug, _inp, _out, _read, _write, _inp_above, _out_above, _read_above, _write_above
 
+# OpenAI Ultrafast (``service_tier: "ultrafast"``): 6x Standard on every bucket, same 272K
+# whole-request tier. Selected by the tier the response reports it was SERVED at (a request asking
+# for Ultrafast can be served at ``default``, and is then billed at Standard).
+_OPENAI_ULTRAFAST_PRICING: Dict[str, PricingEntry] = {
+    "gpt-6-astra": _snap(
+        "60.00", "300.00", "6.00", "75.00",
+        url="https://developers.openai.com/api/docs/pricing?latest-pricing=ultrafast",
+        version="openai-ultrafast-2026-09",
+        tier_threshold_tokens=272_000,
+        input_cost_per_million_above=Decimal("120.00"),
+        output_cost_per_million_above=Decimal("450.00"),
+        cache_read_cost_per_million_above=Decimal("12.00"),
+        cache_write_cost_per_million_above=Decimal("150.00"),
+    ),
+}
+
 # Context-tiered Gemini Pro: above 200k prompt tokens the *_above rates apply to
 # the whole request (see PricingEntry).
 _OFFICIAL_DOCS_PRICING[("google", "gemini-3.1-pro")] = _snap(
@@ -449,6 +465,19 @@ def _lookup_official_docs_pricing(route: BillingRoute) -> Optional[PricingEntry]
     return _OFFICIAL_DOCS_PRICING.get((route.provider, normalized)) if normalized != model else None
 
 
+def with_served_service_tier(usage: CanonicalUsage, response: Any) -> CanonicalUsage:
+    """``usage`` with the response's served ``service_tier`` folded into ``raw_usage``. OpenAI reports
+    the tier on the response, not inside ``usage``, and pricing reads it from ``raw_usage``."""
+    tier = getattr(response, "service_tier", None)
+    if not isinstance(tier, str) or not tier.strip():
+        return usage
+    return replace(usage, raw_usage={**(usage.raw_usage or {}), "service_tier": tier.strip().lower()})
+
+
+def _served_openai_tier(usage: CanonicalUsage) -> Optional[str]:
+    return usage.raw_usage.get("service_tier") if isinstance(usage.raw_usage, dict) else None
+
+
 def _served_fast(usage: CanonicalUsage) -> bool:
     """Anthropic names the speed that served a fast-mode request in ``usage.speed``."""
     return isinstance(usage.raw_usage, dict) and usage.raw_usage.get("speed") == "fast"
@@ -653,6 +682,10 @@ def estimate_usage_cost(
         entry = _anthropic_fast_mode_entry(route.model)
         if not entry:
             return _unknown_cost("official_docs_snapshot", "fast-mode pricing unavailable for model")
+    if route.provider == "openai" and _served_openai_tier(usage) == "ultrafast":
+        entry = _OPENAI_ULTRAFAST_PRICING.get(route.model)
+        if not entry:
+            return _unknown_cost("official_docs_snapshot", "ultrafast pricing unavailable for model")
     if not entry:
         return _unknown_cost("none")
 

apps/desktop/src/app/settings/model-settings.tsx

@@ -84,10 +84,10 @@ export function ModelSettingsSkeleton({ subpage }: Pick<ModelSettingsProps, 'sub
   )
 }
 
-// agent.service_tier stores "fast"/"priority"/"on" for fast; anything else is
-// normal (mirrors tui_gateway _load_service_tier).
+// agent.service_tier stores "fast"/"priority"/"on" for fast and "ultrafast" for OpenAI
+// Ultrafast; anything else is normal (mirrors agent.fast_mode.parse_service_tier).
 const isFastTier = (tier: unknown): boolean =>
-  ['fast', 'priority', 'on'].includes(
+  ['fast', 'priority', 'on', 'ultrafast'].includes(
     String(tier ?? '')
       .trim()
       .toLowerCase()

gateway/run_config_loaders.py

@@ -222,17 +222,10 @@ class GatewayConfigLoadersMixin:
 
     @classmethod
     def _load_service_tier(cls) -> str | None:
-        """``agent.service_tier``: fast/priority/on => "priority"; normal/off => None; None when unset/unknown."""
-        raw = cls._cfg_str("agent", "service_tier")
-        value = raw.lower()
-        if not value or value in {"normal", "default", "standard", "off", "none"}:
-            return None
-        if value in {"fast", "priority", "on"}:
-            return "priority"
-        if value in {"auto", "cold"}:
-            return value
-        logger.warning("Unknown service_tier '%s', ignoring", raw)
-        return None
+        """``agent.service_tier`` parsed like the CLI (``hermes_cli.cli_config_load``); None when unset/unknown."""
+        from hermes_cli.cli_config_load import _parse_service_tier_config
+
+        return _parse_service_tier_config(cls._cfg_str("agent", "service_tier"))
 
     @staticmethod
     def _load_show_reasoning() -> bool:

gateway/run_turn.py

@@ -309,6 +309,7 @@ class GatewayTurnMixin:
         """Effective model/runtime config for one turn. With `/fast` priority on, fast-mode
         ``request_overrides`` are deep-merged OVER the per-provider ones so both reach the model."""
         from gateway.run import _deep_merge_request_overrides
+        from agent.fast_mode import STATIC_TIERS
         from hermes_cli.models import resolve_fast_mode_overrides
         # Tests bind this method onto bare namespaces, so no class-level tables here.
         runtime = {
@@ -328,13 +329,14 @@ class GatewayTurnMixin:
                 runtime["api_mode"], runtime["command"], tuple(runtime["args"]),
             ),
         }
-        if getattr(self, "_service_tier", None) != "priority":
+        tier = getattr(self, "_service_tier", None)
+        if tier not in STATIC_TIERS:
             # None / auto / cold: the bounded window is applied per request by agent.fast_mode.
             route["request_overrides"] = base_request_overrides
             return route
         try:
             overrides = resolve_fast_mode_overrides(
-                route["model"], provider=runtime["provider"], base_url=runtime["base_url"],
+                route["model"], provider=runtime["provider"], base_url=runtime["base_url"], tier=tier,
             )
         except Exception:
             overrides = None

gateway/slash_commands_model.py

@@ -29,6 +29,7 @@ _FAST_SELECTIONS = {
     "off": (None, "normal", "gateway.fast.label_normal"),
     "auto": ("auto", "auto", None),
     "cold": ("cold", "cold", None),
+    "ultrafast": ("ultrafast", "ultrafast", None),
 }
 
 # /reasoning display-toggle arguments -> show_reasoning value.
@@ -804,18 +805,23 @@ class GatewayModelCommandsMixin:
     async def _handle_fast_command(self, event: MessageEvent) -> Optional[str]:
         """Handle /fast — the CLI Priority Processing toggle; session-scoped unless ``--global``
         (persists agent.service_tier, parity with /model)."""
+        from agent.fast_mode import service_tier_word
         from gateway.run import _load_gateway_config, _resolve_gateway_model
-        from hermes_cli.models import model_supports_fast_mode
+        from hermes_cli.models import model_supports_fast_mode, model_supports_ultrafast
 
         # The /reasoning parser strips --global (any position) and normalizes unicode dashes.
         args, persist_global = self._parse_reasoning_command_args(event.get_command_args().strip().lower())
         session_key = self._session_key_for_source(event.source)
         self._service_tier = self._resolve_session_service_tier(session_key=session_key)
-        if not model_supports_fast_mode(_resolve_gateway_model(_load_gateway_config())):
+        model = _resolve_gateway_model(_load_gateway_config())
+        if not model_supports_fast_mode(model):
             return t("gateway.fast.not_supported")
+        ultrafast = model_supports_ultrafast(model)
+        if args == "ultrafast" and not ultrafast:
+            return t("gateway.fast.ultrafast_not_supported", model=model)
         if args and args != "status":
             return self._apply_fast_selection(session_key, args, persist=persist_global)
-        mode = "fast" if self._service_tier == "priority" else (self._service_tier or "normal")
+        mode = service_tier_word(self._service_tier)
         status = {"fast": t("gateway.fast.status_fast"), "normal": t("gateway.fast.status_normal")}.get(mode, mode)
 
         async def _on_fast_choice(_chat_id: str, value: str) -> str:
@@ -827,7 +833,7 @@ class GatewayModelCommandsMixin:
             title=t("gateway.fast.picker_title", mode=status),
             choices=[
                 {"value": v, "label": t(f"gateway.fast.choice_{v}"), "is_current": mode == v}
-                for v in ("fast", "normal", "auto", "cold")
+                for v in ("fast", "normal", "auto", "cold", *(("ultrafast",) if ultrafast else ()))
             ],
             on_choice_selected=_on_fast_choice,
         )

hermes_cli/cli_agent_setup_mixin.py

@@ -523,16 +523,18 @@ class CLIAgentSetupMixin:
 
     def _resolve_turn_agent_config(self, user_message: str) -> dict:
         """Effective model/runtime config for one turn — always the session's primary
-        provider. With `/fast` on (service_tier == "priority") attach request_overrides;
+        provider. With a static `/fast` tier (fast / ultrafast) attach request_overrides;
         auto/cold tiers are applied per request by agent.fast_mode instead."""
+        from agent.fast_mode import STATIC_TIERS
         from hermes_cli.models import resolve_fast_mode_overrides
         runtime = _current_runtime(self)
         route = {"model": self.model, "runtime": runtime, "signature": _route_signature(self.model, runtime)}
         overrides = None
-        if getattr(self, "service_tier", None) == "priority":
+        tier = getattr(self, "service_tier", None)
+        if tier in STATIC_TIERS:
             try:
                 overrides = resolve_fast_mode_overrides(
-                    route["model"], provider=runtime["provider"], base_url=runtime["base_url"])
+                    route["model"], provider=runtime["provider"], base_url=runtime["base_url"], tier=tier)
             except Exception:
                 pass
         route["request_overrides"] = overrides

hermes_cli/cli_commands_mixin.py

@@ -191,7 +191,8 @@ _BUSY_MODES = ("queue", "steer", "interrupt")
 # /fast argument -> (service_tier value, persisted config value)
 _FAST_TIERS = {
     "fast": ("priority", "fast"), "on": ("priority", "fast"), "normal": (None, "normal"),
-    "off": (None, "normal"), "auto": ("auto", "auto"), "cold": ("cold", "cold")}
+    "off": (None, "normal"), "auto": ("auto", "auto"), "cold": ("cold", "cold"),
+    "ultrafast": ("ultrafast", "ultrafast")}
 
 # /reasoning display toggles: arg -> (attr, value, headline key, follow-up note key or None);
 # the keys resolve under ``cli.commands.reasoning.*`` at call time.
@@ -2638,11 +2639,16 @@ class CLICommandsMixin:
         raw = _command_arg(cmd)
         usage = _dim_line(_t("fast.usage"))
         if not raw or raw.lower() == "status":
-            status = {"priority": "fast", None: "normal"}.get(self.service_tier, self.service_tier)
+            from agent.fast_mode import service_tier_word
+            status = service_tier_word(self.service_tier)
             return _cp(_accent_line(_t("fast.status", feature=feature_name, status=status)), usage)
         arg, explicit_global = _split_scope_flags(raw)
         if arg not in _FAST_TIERS:
             return _cp(_dim_line(_t("shared.unknown_argument", arg=arg)), usage)
+        if arg == "ultrafast":
+            if not _probe("hermes_cli.models", "model_supports_ultrafast", False, model):
+                return _cp(_dim_line(_t("fast.ultrafast_not_supported", model=model or "?")), usage)
+            feature_name = _t("fast.feature_ultrafast")
         self.service_tier, saved_value = _FAST_TIERS[arg]
         _retire_agent(self)  # Force agent re-init with new service-tier config
         saved = explicit_global and _save("agent.service_tier", saved_value)

hermes_cli/cli_config_load.py

@@ -69,16 +69,13 @@ def _parse_reasoning_config(effort) -> dict | None:
 
 
 def _parse_service_tier_config(raw: str) -> str | None:
-    """Parse a persisted fast-mode preference: None, "priority", "auto", or "cold"."""
-    value = str(raw or "").strip().lower()
-    if not value or value in {"normal", "default", "standard", "off", "none"}:
-        return None
-    if value in {"fast", "priority", "on"}:
-        return "priority"
-    if value in {"auto", "cold"}:
-        return value
-    logger.warning("Unknown service_tier '%s', ignoring", raw)
-    return None
+    """Parse a persisted fast-mode preference: None, "priority", "ultrafast", "auto", or "cold"."""
+    from agent.fast_mode import NORMAL_TIER_WORDS, parse_service_tier
+
+    tier = parse_service_tier(raw)
+    if tier is None and str(raw or "").strip().lower() not in NORMAL_TIER_WORDS:
+        logger.warning("Unknown service_tier '%s', ignoring", raw)
+    return tier
 
 
 # terminal.<key> -> TERMINAL_<KEY> env var. Container-resource keys apply to docker,

hermes_cli/models.py

@@ -42,6 +42,7 @@ from hermes_cli.models_catalog_static import (
     _LIVE_FIRST_PICKER_PROVIDERS,
     _MODELS_DEV_PREFERRED,
     _OPENAI_FAST_MODE_PREFIXES,
+    _OPENAI_ULTRAFAST_MODELS,
     _PROVIDER_ALIASES,
     _PROVIDER_LABELS,
     _PROVIDER_MODELS,
@@ -1185,17 +1186,30 @@ def _fast_mode_route_supported(
     return not host or host in allowed.values()
 
 
+def model_supports_ultrafast(model_id: Optional[str]) -> bool:
+    """OpenAI Ultrafast (``service_tier: "ultrafast"``) is published per model, not per family."""
+    from agent.model_metadata import strip_codex_context_variant_suffix
+
+    base = _strip_vendor_prefix(strip_codex_context_variant_suffix(str(model_id or ""))).split(":")[0]
+    return base in _OPENAI_ULTRAFAST_MODELS
+
+
 def resolve_fast_mode_overrides(
-    model_id: Optional[str], *, provider: Optional[str] = None, base_url: Optional[str] = None
+    model_id: Optional[str], *, provider: Optional[str] = None, base_url: Optional[str] = None,
+    tier: Optional[str] = None,
 ) -> dict[str, Any] | None:
     """Fast/priority request_overrides — ``{"speed": "fast"}`` (Anthropic Fast Mode) or
     ``{"service_tier": "priority"}`` (OpenAI / xAI Priority Processing) — or None if unsupported.
+    ``tier="ultrafast"`` asks for OpenAI Ultrafast instead: ``{"service_tier": "ultrafast"}`` on an
+    Ultrafast model, None elsewhere (never a silent downgrade to a different paid tier).
     With ``provider``/``base_url`` the route is gated too (``_fast_mode_route_supported``) so proxies
     never see the params. Single fast-mode gate for ``/fast`` and ``agent.fast_mode`` windows."""
     if not model_supports_fast_mode(model_id):
         return None
     if (provider or base_url) and not _fast_mode_route_supported(model_id, provider, base_url):
         return None
+    if tier == "ultrafast":
+        return {"service_tier": "ultrafast"} if model_supports_ultrafast(model_id) else None
     return {"speed": "fast"} if _is_anthropic_fast_model(model_id) else {"service_tier": "priority"}
 
 

hermes_cli/models_catalog_static.py

@@ -560,6 +560,10 @@ _LIVE_FIRST_PICKER_PROVIDERS: frozenset[str] = frozenset({"opencode-zen", "openc
 # positives are harmless. Codex-series models are excluded — the Codex Responses API doesn't
 # expose service_tier.
 _OPENAI_FAST_MODE_PREFIXES: tuple[str, ...] = ("gpt-", "o1", "o3", "o4")
+# OpenAI Ultrafast (service_tier="ultrafast", 6x Standard): broadly available for GPT-6 Astra only
+# (developers.openai.com/api/docs/guides/ultrafast-mode, 2026-09-29); GPT-6.1 Sol "coming soon".
+# Exact wire slugs, matched after stripping the vendor prefix and the Hermes-side ``-900k`` alias.
+_OPENAI_ULTRAFAST_MODELS: frozenset[str] = frozenset({"gpt-6-astra"})
 
 
 # Providers where models.dev is authoritative: the curated list is an offline fallback plus custom

locales/af.tui.yaml

@@ -933,7 +933,7 @@ slashCmd.session.compress.compressedOther: "{0} boodskappe saamgepers"
 slashCmd.session.compress.nothing: "niks om saam te pers nie"
 slashCmd.session.compress.tokSuffix: " · {0} tok"
 slashCmd.session.fast.mode: "vinnige modus: {0}"
-slashCmd.session.fast.usage: "gebruik: /fast [normal|fast|status|on|off|toggle]"
+slashCmd.session.fast.usage: "gebruik: /fast [normal|fast|ultrafast|status|on|off|toggle]"
 slashCmd.session.indicator.current: "aanwyser: {0}"
 slashCmd.session.indicator.switched: "aanwyser → {0}"
 slashCmd.session.indicator.usage: "gebruik: /indicator [{0}]"

locales/af.yaml

@@ -184,6 +184,8 @@ gateway:
     choice_normal: "normal — standaardverwerking"
     choice_auto: "auto — vinnig vir die eerste sekondes van elke beurt"
     choice_cold: "cold — vinnig slegs vir die eerste beurt van 'n sessie"
+    choice_ultrafast: "ultrafast — OpenAI Ultrafast (GPT-6 Astra, 6x prys)"
+    ultrafast_not_supported: "⚡ Ultrafast is slegs beskikbaar vir GPT-6 Astra (huidige model: `{model}`)."
   footer:
     status: "📎 Looptyd-voetstuk: **{state}**\nVelde: `{fields}`\nPlatform: `{platform}`"
     usage: "Gebruik: `/footer [on|off|status]`"
@@ -1598,7 +1600,7 @@ slash:
   reasoning:
     description: "Bestuur redenering-inspanning en -vertoon"
   fast:
-    description: "Vinnige modus — OpenAI Priority Processing / Anthropic Fast Mode (normal/fast/auto/cold)"
+    description: "Vinnige modus — OpenAI Priority Processing / Anthropic Fast Mode (normal/fast/auto/cold/ultrafast)"
   skin:
     description: "Wys of verander die vertoon-skin/tema"
   indicator:
@@ -2651,7 +2653,9 @@ cli:
       feature_generic: "Vinnige modus"
       feature_anthropic: "Anthropic Fast Mode"
       feature_openai: "Priority Processing"
-      usage: "Gebruik: /fast [normal|fast|auto|cold|status] [--global]"
+      feature_ultrafast: "OpenAI Ultrafast"
+      ultrafast_not_supported: "(._.) Ultrafast is slegs beskikbaar vir GPT-6 Astra (huidige model: {model})."
+      usage: "Gebruik: /fast [normal|fast|auto|cold|ultrafast|status] [--global]"
       status: "{feature}: {status}"
       set_to: "✓ {feature} gestel op {value} {scope}"
     hatch:

locales/ar.tui.yaml

@@ -933,7 +933,7 @@ slashCmd.session.compress.compressedOther: "تم ضغط {0} من الرسائل"
 slashCmd.session.compress.nothing: "لا يوجد ما يُضغط"
 slashCmd.session.compress.tokSuffix: " · {0} رمزًا"
 slashCmd.session.fast.mode: "الوضع السريع: {0}"
-slashCmd.session.fast.usage: "الاستخدام: /fast [normal|fast|status|on|off|toggle]"
+slashCmd.session.fast.usage: "الاستخدام: /fast [normal|fast|ultrafast|status|on|off|toggle]"
 slashCmd.session.indicator.current: "المؤشر: {0}"
 slashCmd.session.indicator.switched: "المؤشر → {0}"
 slashCmd.session.indicator.usage: "الاستخدام: /indicator [{0}]"

locales/ar.yaml

@@ -184,6 +184,8 @@ gateway:
     choice_normal: "normal — المعالجة القياسية"
     choice_auto: "auto — سريع في الثواني الأولى من كل دور"
     choice_cold: "cold — سريع في الدور الأول من الجلسة فقط"
+    choice_ultrafast: "ultrafast — ‏OpenAI Ultrafast ‏(GPT-6 Astra، السعر ×6)"
+    ultrafast_not_supported: "⚡ ‏Ultrafast متاح فقط لـ GPT-6 Astra (النموذج الحالي: `{model}`)."
   footer:
     status: "📎 تذييل التشغيل: **{state}**\nالحقول: `{fields}`\nالمنصّة: `{platform}`"
     usage: "الاستخدام: `/footer [on|off|status]`"
@@ -1598,7 +1600,7 @@ slash:
   reasoning:
     description: "إدارة مستوى الاستدلال وطريقة عرضه"
   fast:
-    description: "الوضع السريع — OpenAI Priority Processing / Anthropic Fast Mode (normal/fast/auto/cold)"
+    description: "الوضع السريع — OpenAI Priority Processing / Anthropic Fast Mode (normal/fast/auto/cold/ultrafast)"
   skin:
     description: "عرض سمة العرض أو تغييرها"
   indicator:
@@ -2651,7 +2653,9 @@ cli:
       feature_generic: "الوضع السريع"
       feature_anthropic: "الوضع السريع من Anthropic"
       feature_openai: "المعالجة ذات الأولوية"
-      usage: "الاستخدام: /fast [normal|fast|auto|cold|status] [--global]"
+      feature_ultrafast: "OpenAI Ultrafast"
+      ultrafast_not_supported: "(._.) ‏Ultrafast متاح فقط لـ GPT-6 Astra (النموذج الحالي: {model})."
+      usage: "الاستخدام: /fast [normal|fast|auto|cold|ultrafast|status] [--global]"
       status: "{feature}: {status}"
       set_to: "✓ تم ضبط {feature} على {value} {scope}"
     hatch:

locales/de.tui.yaml

@@ -933,7 +933,7 @@ slashCmd.session.compress.compressedOther: "{0} Nachrichten komprimiert"
 slashCmd.session.compress.nothing: "nichts zu komprimieren"
 slashCmd.session.compress.tokSuffix: " · {0} Tok"
 slashCmd.session.fast.mode: "Fast-Modus: {0}"
-slashCmd.session.fast.usage: "Verwendung: /fast [normal|fast|status|on|off|toggle]"
+slashCmd.session.fast.usage: "Verwendung: /fast [normal|fast|ultrafast|status|on|off|toggle]"
 slashCmd.session.indicator.current: "Indikator: {0}"
 slashCmd.session.indicator.switched: "Indikator → {0}"
 slashCmd.session.indicator.usage: "Verwendung: /indicator [{0}]"

locales/de.yaml

@@ -184,6 +184,8 @@ gateway:
     choice_normal: "normal — Standardverarbeitung"
     choice_auto: "auto — schnell in den ersten Sekunden jedes Zugs"
     choice_cold: "cold — schnell nur im ersten Zug einer Sitzung"
+    choice_ultrafast: "ultrafast — OpenAI Ultrafast (GPT-6 Astra, 6-facher Preis)"
+    ultrafast_not_supported: "⚡ Ultrafast ist nur für GPT-6 Astra verfügbar (aktuelles Modell: `{model}`)."
   footer:
     status: "📎 Laufzeit-Fußzeile: **{state}**\nFelder: `{fields}`\nPlattform: `{platform}`"
     usage: "Verwendung: `/footer [on|off|status]`"
@@ -1598,7 +1600,7 @@ slash:
   reasoning:
     description: "Reasoning-Stärke und -Anzeige verwalten"
   fast:
-    description: "Schnellmodus — OpenAI Priority Processing / Anthropic Fast Mode (normal/fast/auto/cold)"
+    description: "Schnellmodus — OpenAI Priority Processing / Anthropic Fast Mode (normal/fast/auto/cold/ultrafast)"
   skin:
     description: "Anzeige-Skin/Theme anzeigen oder ändern"
   indicator:
@@ -2651,7 +2653,9 @@ cli:
       feature_generic: "Schnellmodus"
       feature_anthropic: "Anthropic Fast Mode"
       feature_openai: "Priority Processing"
-      usage: "Verwendung: /fast [normal|fast|auto|cold|status] [--global]"
+      feature_ultrafast: "OpenAI Ultrafast"
+      ultrafast_not_supported: "(._.) Ultrafast ist nur für GPT-6 Astra verfügbar (aktuelles Modell: {model})."
+      usage: "Verwendung: /fast [normal|fast|auto|cold|ultrafast|status] [--global]"
       status: "{feature}: {status}"
       set_to: "✓ {feature} auf {value} gesetzt {scope}"
     hatch:

locales/en.yaml

@@ -174,8 +174,8 @@ gateway:
     denied_reason_plural: "❌ Commands denied ({count} commands). Reason relayed to the agent: \"{reason}\""
   fast:
     not_supported: "⚡ /fast is only available for OpenAI models that support Priority Processing."
-    status: "⚡ Priority Processing\n\nCurrent mode: `{mode}`\n\n_Usage:_ `/fast <normal|fast|auto|cold|status>`"
-    unknown_arg: "⚠️ Unknown argument: `{arg}`\n\n**Valid options:** normal, fast, auto, cold, status"
+    status: "⚡ Priority Processing\n\nCurrent mode: `{mode}`\n\n_Usage:_ `/fast <normal|fast|auto|cold|ultrafast|status>`"
+    unknown_arg: "⚠️ Unknown argument: `{arg}`\n\n**Valid options:** normal, fast, auto, cold, ultrafast, status"
     saved: "⚡ ✓ Priority Processing: **{label}** (saved to config)\n_(takes effect on next message)_"
     session_only: "⚡ ✓ Priority Processing: **{label}** (this session only)"
     label_fast: "FAST"
@@ -187,6 +187,8 @@ gateway:
     choice_normal: "normal — standard processing"
     choice_auto: "auto — fast for the first seconds of every turn"
     choice_cold: "cold — fast for the first turn of a session only"
+    choice_ultrafast: "ultrafast — OpenAI Ultrafast (GPT-6 Astra, 6x price)"
+    ultrafast_not_supported: "⚡ Ultrafast is only available for GPT-6 Astra (current model: `{model}`)."
   footer:
     status: "📎 Runtime footer: **{state}**\nFields: `{fields}`\nPlatform: `{platform}`"
     usage: "Usage: `/footer [on|off|status]`"
@@ -1669,7 +1671,7 @@ slash:
   reasoning:
     description: "Manage reasoning effort and display"
   fast:
-    description: "Fast mode — OpenAI Priority Processing / Anthropic Fast Mode (normal/fast/auto/cold)"
+    description: "Fast mode — OpenAI Priority Processing / Anthropic Fast Mode (normal/fast/auto/cold/ultrafast)"
   skin:
     description: "Show or change the display skin/theme"
   indicator:
@@ -2751,7 +2753,9 @@ cli:
       feature_generic: "Fast mode"
       feature_anthropic: "Anthropic Fast Mode"
       feature_openai: "Priority Processing"
-      usage: "Usage: /fast [normal|fast|auto|cold|status] [--global]"
+      feature_ultrafast: "OpenAI Ultrafast"
+      ultrafast_not_supported: "(._.) Ultrafast is only available for GPT-6 Astra (current model: {model})."
+      usage: "Usage: /fast [normal|fast|auto|cold|ultrafast|status] [--global]"
       status: "{feature}: {status}"
       set_to: "✓ {feature} set to {value} {scope}"
     hatch:

locales/es.tui.yaml

@@ -933,7 +933,7 @@ slashCmd.session.compress.compressedOther: "{0} mensajes comprimidos"
 slashCmd.session.compress.nothing: "nada que comprimir"
 slashCmd.session.compress.tokSuffix: " · {0} tok"
 slashCmd.session.fast.mode: "modo rápido: {0}"
-slashCmd.session.fast.usage: "uso: /fast [normal|fast|status|on|off|toggle]"
+slashCmd.session.fast.usage: "uso: /fast [normal|fast|ultrafast|status|on|off|toggle]"
 slashCmd.session.indicator.current: "indicador: {0}"
 slashCmd.session.indicator.switched: "indicador → {0}"
 slashCmd.session.indicator.usage: "uso: /indicator [{0}]"

locales/es.yaml

@@ -184,6 +184,8 @@ gateway:
     choice_normal: "normal — procesamiento estándar"
     choice_auto: "auto — rápido en los primeros segundos de cada turno"
     choice_cold: "cold — rápido solo en el primer turno de una sesión"
+    choice_ultrafast: "ultrafast — OpenAI Ultrafast (GPT-6 Astra, precio x6)"
+    ultrafast_not_supported: "⚡ Ultrafast solo está disponible para GPT-6 Astra (modelo actual: `{model}`)."
   footer:
     status: "📎 Pie de ejecución: **{state}**\nCampos: `{fields}`\nPlataforma: `{platform}`"
     usage: "Uso: `/footer [on|off|status]`"
@@ -1598,7 +1600,7 @@ slash:
   reasoning:
     description: "Gestionar el esfuerzo de razonamiento y su visualización"
   fast:
-    description: "Modo rápido: OpenAI Priority Processing / Anthropic Fast Mode (normal/fast/auto/cold)"
+    description: "Modo rápido: OpenAI Priority Processing / Anthropic Fast Mode (normal/fast/auto/cold/ultrafast)"
   skin:
     description: "Mostrar o cambiar el skin/tema de la interfaz"
   indicator:
@@ -2651,7 +2653,9 @@ cli:
       feature_generic: "Modo rápido"
       feature_anthropic: "Anthropic Fast Mode"
       feature_openai: "Priority Processing"
-      usage: "Uso: /fast [normal|fast|auto|cold|status] [--global]"
+      feature_ultrafast: "OpenAI Ultrafast"
+      ultrafast_not_supported: "(._.) Ultrafast solo está disponible para GPT-6 Astra (modelo actual: {model})."
+      usage: "Uso: /fast [normal|fast|auto|cold|ultrafast|status] [--global]"
       status: "{feature}: {status}"
       set_to: "✓ {feature} establecido en {value} {scope}"
     hatch:

locales/fr.tui.yaml

@@ -933,7 +933,7 @@ slashCmd.session.compress.compressedOther: "{0} messages compressés"
 slashCmd.session.compress.nothing: "rien à compresser"
 slashCmd.session.compress.tokSuffix: " · {0} tok"
 slashCmd.session.fast.mode: "mode rapide : {0}"
-slashCmd.session.fast.usage: "usage : /fast [normal|fast|status|on|off|toggle]"
+slashCmd.session.fast.usage: "usage : /fast [normal|fast|ultrafast|status|on|off|toggle]"
 slashCmd.session.indicator.current: "indicateur : {0}"
 slashCmd.session.indicator.switched: "indicateur → {0}"
 slashCmd.session.indicator.usage: "usage : /indicator [{0}]"

locales/fr.yaml

@@ -210,6 +210,8 @@ gateway:
     choice_normal: "normal — traitement standard"
     choice_auto: "auto — rapide pendant les premières secondes de chaque tour"
     choice_cold: "cold — rapide uniquement au premier tour d'une session"
+    choice_ultrafast: "ultrafast — OpenAI Ultrafast (GPT-6 Astra, prix x6)"
+    ultrafast_not_supported: "⚡ Ultrafast n'est disponible que pour GPT-6 Astra (modèle actuel : `{model}`)."
   footer:
     status: "📎 Pied de page d'exécution : **{state}**\nChamps : `{fields}`\nPlateforme : `{platform}`"
     usage: "Usage : `/footer [on|off|status]`"
@@ -1850,7 +1852,7 @@ slash:
   reasoning:
     description: "Gérer l'effort de raisonnement et son affichage"
   fast:
-    description: "Mode rapide — OpenAI Priority Processing / Anthropic Fast Mode (normal/fast/auto/cold)"
+    description: "Mode rapide — OpenAI Priority Processing / Anthropic Fast Mode (normal/fast/auto/cold/ultrafast)"
   skin:
     description: "Afficher ou changer le skin/thème d'affichage"
   indicator:
@@ -2929,7 +2931,9 @@ cli:
       feature_generic: "Mode rapide"
       feature_anthropic: "Anthropic Fast Mode"
       feature_openai: "Priority Processing"
-      usage: "Usage : /fast [normal|fast|auto|cold|status] [--global]"
+      feature_ultrafast: "OpenAI Ultrafast"
+      ultrafast_not_supported: "(._.) Ultrafast n'est disponible que pour GPT-6 Astra (modèle actuel : {model})."
+      usage: "Usage : /fast [normal|fast|auto|cold|ultrafast|status] [--global]"
       status: "{feature} : {status}"
       set_to: "✓ {feature} défini sur {value} {scope}"
     hatch:

locales/ga.tui.yaml

@@ -933,7 +933,7 @@ slashCmd.session.compress.compressedOther: "dlúthaíodh {0} teachtaireacht"
 slashCmd.session.compress.nothing: "níl aon rud le dlúthú"
 slashCmd.session.compress.tokSuffix: " · {0} comhartha"
 slashCmd.session.fast.mode: "mód tapa: {0}"
-slashCmd.session.fast.usage: "úsáid: /fast [normal|fast|status|on|off|toggle]"
+slashCmd.session.fast.usage: "úsáid: /fast [normal|fast|ultrafast|status|on|off|toggle]"
 slashCmd.session.indicator.current: "táscaire: {0}"
 slashCmd.session.indicator.switched: "táscaire → {0}"
 slashCmd.session.indicator.usage: "úsáid: /indicator [{0}]"

locales/ga.yaml

@@ -184,6 +184,8 @@ gateway:
     choice_normal: "normal — gnáthphróiseáil"
     choice_auto: "auto — tapa do na chéad soicindí de gach seal"
     choice_cold: "cold — tapa don chéad seal de sheisiún amháin"
+    choice_ultrafast: "ultrafast — OpenAI Ultrafast (GPT-6 Astra, praghas x6)"
+    ultrafast_not_supported: "⚡ Níl Ultrafast ar fáil ach do GPT-6 Astra (samhail reatha: `{model}`)."
   footer:
     status: "📎 Buntásc rite: **{state}**\nRéimsí: `{fields}`\nArdán: `{platform}`"
     usage: "Úsáid: `/footer [on|off|status]`"
@@ -1598,7 +1600,7 @@ slash:
   reasoning:
     description: "Bainistigh iarracht agus taispeáint na réasúnaíochta"
   fast:
-    description: "Mód tapa — OpenAI Priority Processing / Anthropic Fast Mode (normal/fast/auto/cold)"
+    description: "Mód tapa — OpenAI Priority Processing / Anthropic Fast Mode (normal/fast/auto/cold/ultrafast)"
   skin:
     description: "Taispeáin nó athraigh craiceann/téama na taispeána"
   indicator:
@@ -2651,7 +2653,9 @@ cli:
       feature_generic: "Mód tapa"
       feature_anthropic: "Anthropic Fast Mode"
       feature_openai: "Priority Processing"
-      usage: "Úsáid: /fast [normal|fast|auto|cold|status] [--global]"
+      feature_ultrafast: "OpenAI Ultrafast"
+      ultrafast_not_supported: "(._.) Níl Ultrafast ar fáil ach do GPT-6 Astra (samhail reatha: {model})."
+      usage: "Úsáid: /fast [normal|fast|auto|cold|ultrafast|status] [--global]"
       status: "{feature}: {status}"
       set_to: "✓ {feature} socraithe go {value} {scope}"
     hatch:

locales/hu.tui.yaml

@@ -933,7 +933,7 @@ slashCmd.session.compress.compressedOther: "{0} üzenet tömörítve"
 slashCmd.session.compress.nothing: "nincs mit tömöríteni"
 slashCmd.session.compress.tokSuffix: " · {0} tok"
 slashCmd.session.fast.mode: "gyors mód: {0}"
-slashCmd.session.fast.usage: "használat: /fast [normal|fast|status|on|off|toggle]"
+slashCmd.session.fast.usage: "használat: /fast [normal|fast|ultrafast|status|on|off|toggle]"
 slashCmd.session.indicator.current: "indikátor: {0}"
 slashCmd.session.indicator.switched: "indikátor → {0}"
 slashCmd.session.indicator.usage: "használat: /indicator [{0}]"

locales/hu.yaml

@@ -187,6 +187,8 @@ gateway:
     choice_normal: "normal — normál feldolgozás"
     choice_auto: "auto — gyors minden kör első másodperceiben"
     choice_cold: "cold — gyors csak a munkamenet első körében"
+    choice_ultrafast: "ultrafast — OpenAI Ultrafast (GPT-6 Astra, hatszoros ár)"
+    ultrafast_not_supported: "⚡ Az Ultrafast csak a GPT-6 Astra modellhez érhető el (jelenlegi modell: `{model}`)."
   footer:
     status: "📎 Futási idejű lábléc: **{state}**\nMezők: `{fields}`\nPlatform: `{platform}`"
     usage: "Használat: `/footer [on|off|status]`"
@@ -1615,7 +1617,7 @@ slash:
   reasoning:
     description: "Gondolkodási erőfeszítés és megjelenítés kezelése"
   fast:
-    description: "Gyors mód — OpenAI Priority Processing / Anthropic Fast Mode (normal/fast/auto/cold)"
+    description: "Gyors mód — OpenAI Priority Processing / Anthropic Fast Mode (normal/fast/auto/cold/ultrafast)"
   skin:
     description: "A megjelenítési skin/téma megjelenítése vagy módosítása"
   indicator:
@@ -2668,7 +2670,9 @@ cli:
       feature_generic: "Gyors mód"
       feature_anthropic: "Anthropic Fast Mode"
       feature_openai: "Priority Processing"
-      usage: "Használat: /fast [normal|fast|auto|cold|status] [--global]"
+      feature_ultrafast: "OpenAI Ultrafast"
+      ultrafast_not_supported: "(._.) Az Ultrafast csak a GPT-6 Astra modellhez érhető el (jelenlegi modell: {model})."
+      usage: "Használat: /fast [normal|fast|auto|cold|ultrafast|status] [--global]"
       status: "{feature}: {status}"
       set_to: "✓ {feature} beállítva: {value} {scope}"
     hatch:

locales/it.tui.yaml

@@ -933,7 +933,7 @@ slashCmd.session.compress.compressedOther: "compressi {0} messaggi"
 slashCmd.session.compress.nothing: "niente da comprimere"
 slashCmd.session.compress.tokSuffix: " · {0} tok"
 slashCmd.session.fast.mode: "modalità veloce: {0}"
-slashCmd.session.fast.usage: "uso: /fast [normal|fast|status|on|off|toggle]"
+slashCmd.session.fast.usage: "uso: /fast [normal|fast|ultrafast|status|on|off|toggle]"
 slashCmd.session.indicator.current: "indicatore: {0}"
 slashCmd.session.indicator.switched: "indicatore → {0}"
 slashCmd.session.indicator.usage: "uso: /indicator [{0}]"

locales/it.yaml

@@ -184,6 +184,8 @@ gateway:
     choice_normal: "normal — elaborazione standard"
     choice_auto: "auto — veloce nei primi secondi di ogni turno"
     choice_cold: "cold — veloce solo nel primo turno di una sessione"
+    choice_ultrafast: "ultrafast — OpenAI Ultrafast (GPT-6 Astra, prezzo x6)"
+    ultrafast_not_supported: "⚡ Ultrafast è disponibile solo per GPT-6 Astra (modello attuale: `{model}`)."
   footer:
     status: "📎 Footer di runtime: **{state}**\nCampi: `{fields}`\nPiattaforma: `{platform}`"
     usage: "Uso: `/footer [on|off|status]`"
@@ -1598,7 +1600,7 @@ slash:
   reasoning:
     description: "Gestisci lo sforzo di ragionamento e la sua visualizzazione"
   fast:
-    description: "Modalità veloce — OpenAI Priority Processing / Anthropic Fast Mode (normal/fast/auto/cold)"
+    description: "Modalità veloce — OpenAI Priority Processing / Anthropic Fast Mode (normal/fast/auto/cold/ultrafast)"
   skin:
     description: "Mostra o cambia la skin/tema di visualizzazione"
   indicator:
@@ -2651,7 +2653,9 @@ cli:
       feature_generic: "Modalità veloce"
       feature_anthropic: "Anthropic Fast Mode"
       feature_openai: "Priority Processing"
-      usage: "Uso: /fast [normal|fast|auto|cold|status] [--global]"
+      feature_ultrafast: "OpenAI Ultrafast"
+      ultrafast_not_supported: "(._.) Ultrafast è disponibile solo per GPT-6 Astra (modello attuale: {model})."
+      usage: "Uso: /fast [normal|fast|auto|cold|ultrafast|status] [--global]"
       status: "{feature}: {status}"
       set_to: "✓ {feature} impostato su {value} {scope}"
     hatch:

locales/ja.tui.yaml

@@ -933,7 +933,7 @@ slashCmd.session.compress.compressedOther: "{0} 件のメッセージを圧縮
 slashCmd.session.compress.nothing: "圧縮するものがありません"
 slashCmd.session.compress.tokSuffix: " · {0} tok"
 slashCmd.session.fast.mode: "高速モード: {0}"
-slashCmd.session.fast.usage: "使い方: /fast [normal|fast|status|on|off|toggle]"
+slashCmd.session.fast.usage: "使い方: /fast [normal|fast|ultrafast|status|on|off|toggle]"
 slashCmd.session.indicator.current: "インジケーター: {0}"
 slashCmd.session.indicator.switched: "インジケーター → {0}"
 slashCmd.session.indicator.usage: "使い方: /indicator [{0}]"

locales/ja.yaml

@@ -184,6 +184,8 @@ gateway:
     choice_normal: "normal — 標準処理"
     choice_auto: "auto — 各ターンの最初の数秒間だけ高速"
     choice_cold: "cold — セッションの最初のターンのみ高速"
+    choice_ultrafast: "ultrafast — OpenAI Ultrafast（GPT-6 Astra、料金 6 倍）"
+    ultrafast_not_supported: "⚡ Ultrafast は GPT-6 Astra でのみ利用できます（現在のモデル: `{model}`）。"
   footer:
     status: "📎 ランタイムフッター: **{state}**\nフィールド: `{fields}`\nプラットフォーム: `{platform}`"
     usage: "使い方: `/footer [on|off|status]`"
@@ -2651,7 +2653,9 @@ cli:
       feature_generic: "高速モード"
       feature_anthropic: "Anthropic Fast Mode"
       feature_openai: "Priority Processing"
-      usage: "使い方: /fast [normal|fast|auto|cold|status] [--global]"
+      feature_ultrafast: "OpenAI Ultrafast"
+      ultrafast_not_supported: "(._.) Ultrafast は GPT-6 Astra でのみ利用できます（現在のモデル: {model}）。"
+      usage: "使い方: /fast [normal|fast|auto|cold|ultrafast|status] [--global]"
       status: "{feature}: {status}"
       set_to: "✓ {feature} を {value} に設定しました {scope}"
     hatch:

locales/ko.tui.yaml

@@ -933,7 +933,7 @@ slashCmd.session.compress.compressedOther: "메시지 {0}개 압축함"
 slashCmd.session.compress.nothing: "압축할 내용이 없어요"
 slashCmd.session.compress.tokSuffix: " · {0} 토큰"
 slashCmd.session.fast.mode: "빠른 모드: {0}"
-slashCmd.session.fast.usage: "사용법: /fast [normal|fast|status|on|off|toggle]"
+slashCmd.session.fast.usage: "사용법: /fast [normal|fast|ultrafast|status|on|off|toggle]"
 slashCmd.session.indicator.current: "인디케이터: {0}"
 slashCmd.session.indicator.switched: "인디케이터 → {0}"
 slashCmd.session.indicator.usage: "사용법: /indicator [{0}]"

locales/ko.yaml

@@ -184,6 +184,8 @@ gateway:
     choice_normal: "normal — 표준 처리"
     choice_auto: "auto — 매 턴의 처음 몇 초 동안 빠름"
     choice_cold: "cold — 세션의 첫 턴에만 빠름"
+    choice_ultrafast: "ultrafast — OpenAI Ultrafast (GPT-6 Astra, 가격 6배)"
+    ultrafast_not_supported: "⚡ Ultrafast는 GPT-6 Astra에서만 사용할 수 있습니다 (현재 모델: `{model}`)."
   footer:
     status: "📎 런타임 푸터: **{state}**\n필드: `{fields}`\n플랫폼: `{platform}`"
     usage: "사용법: `/footer [on|off|status]`"
@@ -1598,7 +1600,7 @@ slash:
   reasoning:
     description: "추론 강도와 표시 관리"
   fast:
-    description: "빠른 모드 — OpenAI Priority Processing / Anthropic Fast Mode (normal/fast/auto/cold)"
+    description: "빠른 모드 — OpenAI Priority Processing / Anthropic Fast Mode (normal/fast/auto/cold/ultrafast)"
   skin:
     description: "화면 스킨/테마 표시 또는 변경"
   indicator:
@@ -2651,7 +2653,9 @@ cli:
       feature_generic: "빠른 모드"
       feature_anthropic: "Anthropic Fast Mode"
       feature_openai: "Priority Processing"
-      usage: "사용법: /fast [normal|fast|auto|cold|status] [--global]"
+      feature_ultrafast: "OpenAI Ultrafast"
+      ultrafast_not_supported: "(._.) Ultrafast는 GPT-6 Astra에서만 사용할 수 있습니다 (현재 모델: {model})."
+      usage: "사용법: /fast [normal|fast|auto|cold|ultrafast|status] [--global]"
       status: "{feature}: {status}"
       set_to: "✓ {feature}을(를) {value}(으)로 설정했어요 {scope}"
     hatch:

locales/pt.tui.yaml

@@ -933,7 +933,7 @@ slashCmd.session.compress.compressedOther: "{0} mensagens compactadas"
 slashCmd.session.compress.nothing: "nada para compactar"
 slashCmd.session.compress.tokSuffix: " · {0} tok"
 slashCmd.session.fast.mode: "modo rápido: {0}"
-slashCmd.session.fast.usage: "uso: /fast [normal|fast|status|on|off|toggle]"
+slashCmd.session.fast.usage: "uso: /fast [normal|fast|ultrafast|status|on|off|toggle]"
 slashCmd.session.indicator.current: "indicador: {0}"
 slashCmd.session.indicator.switched: "indicador → {0}"
 slashCmd.session.indicator.usage: "uso: /indicator [{0}]"

locales/pt.yaml

@@ -184,6 +184,8 @@ gateway:
     choice_normal: "normal — processamento padrão"
     choice_auto: "auto — rápido nos primeiros segundos de cada turno"
     choice_cold: "cold — rápido apenas no primeiro turno de uma sessão"
+    choice_ultrafast: "ultrafast — OpenAI Ultrafast (GPT-6 Astra, preço 6x)"
+    ultrafast_not_supported: "⚡ O Ultrafast só está disponível para o GPT-6 Astra (modelo atual: `{model}`)."
   footer:
     status: "📎 Rodapé de execução: **{state}**\nCampos: `{fields}`\nPlataforma: `{platform}`"
     usage: "Uso: `/footer [on|off|status]`"
@@ -1598,7 +1600,7 @@ slash:
   reasoning:
     description: "Gerenciar o esforço e a exibição do raciocínio"
   fast:
-    description: "Modo rápido — OpenAI Priority Processing / Anthropic Fast Mode (normal/fast/auto/cold)"
+    description: "Modo rápido — OpenAI Priority Processing / Anthropic Fast Mode (normal/fast/auto/cold/ultrafast)"
   skin:
     description: "Mostrar ou alterar a skin/tema de exibição"
   indicator:
@@ -2651,7 +2653,9 @@ cli:
       feature_generic: "Modo rápido"
       feature_anthropic: "Anthropic Fast Mode"
       feature_openai: "Priority Processing"
-      usage: "Uso: /fast [normal|fast|auto|cold|status] [--global]"
+      feature_ultrafast: "OpenAI Ultrafast"
+      ultrafast_not_supported: "(._.) O Ultrafast só está disponível para o GPT-6 Astra (modelo atual: {model})."
+      usage: "Uso: /fast [normal|fast|auto|cold|ultrafast|status] [--global]"
       status: "{feature}: {status}"
       set_to: "✓ {feature} definido como {value} {scope}"
     hatch:

locales/ru.tui.yaml

@@ -933,7 +933,7 @@ slashCmd.session.compress.compressedOther: "сжато сообщений: {0}"
 slashCmd.session.compress.nothing: "нечего сжимать"
 slashCmd.session.compress.tokSuffix: " · {0} ток."
 slashCmd.session.fast.mode: "быстрый режим: {0}"
-slashCmd.session.fast.usage: "Использование: /fast [normal|fast|status|on|off|toggle]"
+slashCmd.session.fast.usage: "Использование: /fast [normal|fast|ultrafast|status|on|off|toggle]"
 slashCmd.session.indicator.current: "индикатор: {0}"
 slashCmd.session.indicator.switched: "индикатор → {0}"
 slashCmd.session.indicator.usage: "Использование: /indicator [{0}]"

locales/ru.yaml

@@ -184,6 +184,8 @@ gateway:
     choice_normal: "normal — стандартная обработка"
     choice_auto: "auto — быстро в первые секунды каждого хода"
     choice_cold: "cold — быстро только на первом ходе сессии"
+    choice_ultrafast: "ultrafast — OpenAI Ultrafast (GPT-6 Astra, цена ×6)"
+    ultrafast_not_supported: "⚡ Ultrafast доступен только для GPT-6 Astra (текущая модель: `{model}`)."
   footer:
     status: "📎 Нижний колонтитул среды выполнения: **{state}**\nПоля: `{fields}`\nПлатформа: `{platform}`"
     usage: "Использование: `/footer [on|off|status]`"
@@ -1598,7 +1600,7 @@ slash:
   reasoning:
     description: "Управлять уровнем и отображением рассуждений"
   fast:
-    description: "Быстрый режим — OpenAI Priority Processing / Anthropic Fast Mode (normal/fast/auto/cold)"
+    description: "Быстрый режим — OpenAI Priority Processing / Anthropic Fast Mode (normal/fast/auto/cold/ultrafast)"
   skin:
     description: "Показать или сменить скин/тему оформления"
   indicator:
@@ -2651,7 +2653,9 @@ cli:
       feature_generic: "Быстрый режим"
       feature_anthropic: "Anthropic Fast Mode"
       feature_openai: "Priority Processing"
-      usage: "Использование: /fast [normal|fast|auto|cold|status] [--global]"
+      feature_ultrafast: "OpenAI Ultrafast"
+      ultrafast_not_supported: "(._.) Ultrafast доступен только для GPT-6 Astra (текущая модель: {model})."
+      usage: "Использование: /fast [normal|fast|auto|cold|ultrafast|status] [--global]"
       status: "{feature}: {status}"
       set_to: "✓ {feature}: установлено {value} {scope}"
     hatch:

locales/tr.tui.yaml

@@ -933,7 +933,7 @@ slashCmd.session.compress.compressedOther: "{0} mesaj sıkıştırıldı"
 slashCmd.session.compress.nothing: "sıkıştırılacak bir şey yok"
 slashCmd.session.compress.tokSuffix: " · {0} tok"
 slashCmd.session.fast.mode: "hızlı mod: {0}"
-slashCmd.session.fast.usage: "kullanım: /fast [normal|fast|status|on|off|toggle]"
+slashCmd.session.fast.usage: "kullanım: /fast [normal|fast|ultrafast|status|on|off|toggle]"
 slashCmd.session.indicator.current: "gösterge: {0}"
 slashCmd.session.indicator.switched: "gösterge → {0}"
 slashCmd.session.indicator.usage: "kullanım: /indicator [{0}]"

locales/tr.yaml

@@ -184,6 +184,8 @@ gateway:
     choice_normal: "normal — standart işleme"
     choice_auto: "auto — her turun ilk saniyelerinde hızlı"
     choice_cold: "cold — yalnızca oturumun ilk turunda hızlı"
+    choice_ultrafast: "ultrafast — OpenAI Ultrafast (GPT-6 Astra, 6 kat fiyat)"
+    ultrafast_not_supported: "⚡ Ultrafast yalnızca GPT-6 Astra için kullanılabilir (geçerli model: `{model}`)."
   footer:
     status: "📎 Çalışma zamanı altbilgisi: **{state}**\nAlanlar: `{fields}`\nPlatform: `{platform}`"
     usage: "Kullanım: `/footer [on|off|status]`"
@@ -1598,7 +1600,7 @@ slash:
   reasoning:
     description: "Akıl yürütme düzeyini ve gösterimini yönet"
   fast:
-    description: "Hızlı mod — OpenAI Priority Processing / Anthropic Fast Mode (normal/fast/auto/cold)"
+    description: "Hızlı mod — OpenAI Priority Processing / Anthropic Fast Mode (normal/fast/auto/cold/ultrafast)"
   skin:
     description: "Görüntü görünümünü/temasını göster veya değiştir"
   indicator:
@@ -2651,7 +2653,9 @@ cli:
       feature_generic: "Hızlı mod"
       feature_anthropic: "Anthropic Fast Mode"
       feature_openai: "Priority Processing"
-      usage: "Kullanım: /fast [normal|fast|auto|cold|status] [--global]"
+      feature_ultrafast: "OpenAI Ultrafast"
+      ultrafast_not_supported: "(._.) Ultrafast yalnızca GPT-6 Astra için kullanılabilir (geçerli model: {model})."
+      usage: "Kullanım: /fast [normal|fast|auto|cold|ultrafast|status] [--global]"
       status: "{feature}: {status}"
       set_to: "✓ {feature} {value} olarak ayarlandı {scope}"
     hatch:

locales/uk.tui.yaml

@@ -933,7 +933,7 @@ slashCmd.session.compress.compressedOther: "стиснуто повідомле
 slashCmd.session.compress.nothing: "нічого стискати"
 slashCmd.session.compress.tokSuffix: " · {0} ток"
 slashCmd.session.fast.mode: "швидкий режим: {0}"
-slashCmd.session.fast.usage: "використання: /fast [normal|fast|status|on|off|toggle]"
+slashCmd.session.fast.usage: "використання: /fast [normal|fast|ultrafast|status|on|off|toggle]"
 slashCmd.session.indicator.current: "індикатор: {0}"
 slashCmd.session.indicator.switched: "індикатор → {0}"
 slashCmd.session.indicator.usage: "використання: /indicator [{0}]"

locales/uk.yaml

@@ -184,6 +184,8 @@ gateway:
     choice_normal: "normal — стандартна обробка"
     choice_auto: "auto — швидко в перші секунди кожного ходу"
     choice_cold: "cold — швидко лише на першому ході сесії"
+    choice_ultrafast: "ultrafast — OpenAI Ultrafast (GPT-6 Astra, ціна ×6)"
+    ultrafast_not_supported: "⚡ Ultrafast доступний лише для GPT-6 Astra (поточна модель: `{model}`)."
   footer:
     status: "📎 Нижній колонтитул середовища: **{state}**\nПоля: `{fields}`\nПлатформа: `{platform}`"
     usage: "Використання: `/footer [on|off|status]`"
@@ -1598,7 +1600,7 @@ slash:
   reasoning:
     description: "Керувати рівнем зусиль міркування та його показом"
   fast:
-    description: "Швидкий режим — OpenAI Priority Processing / Anthropic Fast Mode (normal/fast/auto/cold)"
+    description: "Швидкий режим — OpenAI Priority Processing / Anthropic Fast Mode (normal/fast/auto/cold/ultrafast)"
   skin:
     description: "Показати або змінити оформлення (тему) інтерфейсу"
   indicator:
@@ -2651,7 +2653,9 @@ cli:
       feature_generic: "Швидкий режим"
       feature_anthropic: "Anthropic Fast Mode"
       feature_openai: "Priority Processing"
-      usage: "Використання: /fast [normal|fast|auto|cold|status] [--global]"
+      feature_ultrafast: "OpenAI Ultrafast"
+      ultrafast_not_supported: "(._.) Ultrafast доступний лише для GPT-6 Astra (поточна модель: {model})."
+      usage: "Використання: /fast [normal|fast|auto|cold|ultrafast|status] [--global]"
       status: "{feature}: {status}"
       set_to: "✓ {feature}: встановлено {value} {scope}"
     hatch:

locales/zh-hant.tui.yaml

@@ -933,7 +933,7 @@ slashCmd.session.compress.compressedOther: "已壓縮 {0} 則訊息"
 slashCmd.session.compress.nothing: "沒有可壓縮的內容"
 slashCmd.session.compress.tokSuffix: " · {0} tok"
 slashCmd.session.fast.mode: "快速模式：{0}"
-slashCmd.session.fast.usage: "用法：/fast [normal|fast|status|on|off|toggle]"
+slashCmd.session.fast.usage: "用法：/fast [normal|fast|ultrafast|status|on|off|toggle]"
 slashCmd.session.indicator.current: "指示器：{0}"
 slashCmd.session.indicator.switched: "指示器 → {0}"
 slashCmd.session.indicator.usage: "用法：/indicator [{0}]"

locales/zh-hant.yaml

@@ -184,6 +184,8 @@ gateway:
     choice_normal: "normal — 標準處理"
     choice_auto: "auto — 每輪的前幾秒快速"
     choice_cold: "cold — 僅會話的第一輪快速"
+    choice_ultrafast: "ultrafast — OpenAI Ultrafast（GPT-6 Astra，6 倍價格）"
+    ultrafast_not_supported: "⚡ Ultrafast 僅適用於 GPT-6 Astra（目前模型：`{model}`）。"
   footer:
     status: "📎 執行階段頁尾：**{state}**\n欄位：`{fields}`\n平台：`{platform}`"
     usage: "用法：`/footer [on|off|status]`"
@@ -2651,7 +2653,9 @@ cli:
       feature_generic: "快速模式"
       feature_anthropic: "Anthropic Fast Mode"
       feature_openai: "Priority Processing"
-      usage: "用法：/fast [normal|fast|auto|cold|status] [--global]"
+      feature_ultrafast: "OpenAI Ultrafast"
+      ultrafast_not_supported: "(._.) Ultrafast 僅適用於 GPT-6 Astra（目前模型：{model}）。"
+      usage: "用法：/fast [normal|fast|auto|cold|ultrafast|status] [--global]"
       status: "{feature}：{status}"
       set_to: "✓ {feature} 已設為 {value} {scope}"
     hatch:

locales/zh.tui.yaml

@@ -933,7 +933,7 @@ slashCmd.session.compress.compressedOther: "已压缩 {0} 条消息"
 slashCmd.session.compress.nothing: "没有可压缩的内容"
 slashCmd.session.compress.tokSuffix: " · {0} tok"
 slashCmd.session.fast.mode: "快速模式：{0}"
-slashCmd.session.fast.usage: "用法：/fast [normal|fast|status|on|off|toggle]"
+slashCmd.session.fast.usage: "用法：/fast [normal|fast|ultrafast|status|on|off|toggle]"
 slashCmd.session.indicator.current: "指示器：{0}"
 slashCmd.session.indicator.switched: "指示器 → {0}"
 slashCmd.session.indicator.usage: "用法：/indicator [{0}]"

locales/zh.yaml

@@ -184,6 +184,8 @@ gateway:
     choice_normal: "normal — 标准处理"
     choice_auto: "auto — 每轮的前几秒快速"
     choice_cold: "cold — 仅会话的第一轮快速"
+    choice_ultrafast: "ultrafast — OpenAI Ultrafast（GPT-6 Astra，6 倍价格）"
+    ultrafast_not_supported: "⚡ Ultrafast 仅适用于 GPT-6 Astra（当前模型：`{model}`）。"
   footer:
     status: "📎 运行时页脚：**{state}**\n字段：`{fields}`\n平台：`{platform}`"
     usage: "用法：`/footer [on|off|status]`"
@@ -2651,7 +2653,9 @@ cli:
       feature_generic: "快速模式"
       feature_anthropic: "Anthropic Fast Mode"
       feature_openai: "Priority Processing"
-      usage: "用法：/fast [normal|fast|auto|cold|status] [--global]"
+      feature_ultrafast: "OpenAI Ultrafast"
+      ultrafast_not_supported: "(._.) Ultrafast 仅适用于 GPT-6 Astra（当前模型：{model}）。"
+      usage: "用法：/fast [normal|fast|auto|cold|ultrafast|status] [--global]"
       status: "{feature}：{status}"
       set_to: "✓ {feature} 已设为 {value} {scope}"
     hatch:

tests/agent/test_openai_ultrafast.py

@@ -0,0 +1,106 @@
+"""OpenAI Ultrafast (``service_tier: "ultrafast"``): one tier word table, a per-model request gate,
+and pricing from the tier the response was SERVED at. Relationship tests, no catalog snapshots."""
+
+from types import SimpleNamespace
+
+import pytest
+
+from agent.fast_mode import SERVICE_TIER_WORDS, STATIC_TIERS, parse_service_tier
+from agent.usage_pricing import (
+    _OPENAI_ULTRAFAST_PRICING,
+    CanonicalUsage,
+    estimate_usage_cost,
+    with_served_service_tier,
+)
+from hermes_cli.models import model_supports_ultrafast, resolve_fast_mode_overrides
+
+ASTRA_SPELLINGS = ("gpt-6-astra", "openai/gpt-6-astra", "gpt-6-astra-900k")
+
+
+def test_every_config_loader_parses_tiers_through_the_same_table(monkeypatch):
+    from gateway.run import GatewayRunner
+    from hermes_cli.cli_config_load import _parse_service_tier_config
+    import tui_gateway.server as tui
+
+    for word, tier in {**SERVICE_TIER_WORDS, "normal": None, "off": None, "bogus": None}.items():
+        monkeypatch.setattr(GatewayRunner, "_cfg_str", classmethod(lambda cls, *_k, _w=word: _w))
+        monkeypatch.setattr(tui, "_load_cfg", lambda _w=word: {"agent": {"service_tier": _w}})
+        assert parse_service_tier(word) == tier
+        assert _parse_service_tier_config(word) == tier, word
+        assert GatewayRunner._load_service_tier() == tier, word
+        assert tui._load_service_tier() == tier, word
+    assert "ultrafast" in STATIC_TIERS
+
+
+@pytest.mark.parametrize("provider,base_url", [("openai", "https://api.openai.com/v1"),
+                                               ("openai-codex", "https://chatgpt.com/backend-api/codex")])
+def test_ultrafast_is_requested_only_for_ultrafast_models_on_first_party_routes(provider, base_url):
+    for model in ASTRA_SPELLINGS:
+        assert model_supports_ultrafast(model), model
+        assert resolve_fast_mode_overrides(model, provider=provider, base_url=base_url, tier="ultrafast") == {
+            "service_tier": "ultrafast"}
+    # A Priority-capable model without Ultrafast gets nothing, never a silent swap to another paid tier.
+    assert resolve_fast_mode_overrides("gpt-6-sol", provider=provider, base_url=base_url) == {"service_tier": "priority"}
+    assert resolve_fast_mode_overrides("gpt-6-sol", provider=provider, base_url=base_url, tier="ultrafast") is None
+    # Proxies never see the tier.
+    assert resolve_fast_mode_overrides("openai/gpt-6-astra", provider="openrouter",
+                                       base_url="https://openrouter.ai/api/v1", tier="ultrafast") is None
+
+
+def test_cli_and_gateway_turn_routes_send_the_static_tier():
+    import cli as cli_mod
+    from gateway.run import GatewayRunner
+
+    stub = SimpleNamespace(model="gpt-6-astra", api_key="k", base_url="https://api.openai.com/v1", provider="openai",
+                           api_mode="codex_responses", acp_command=None, acp_args=[], _credential_pool=None,
+                           service_tier="ultrafast")
+    assert cli_mod.HermesCLI._resolve_turn_agent_config(stub, "hi")["request_overrides"] == {"service_tier": "ultrafast"}
+    runner = object.__new__(GatewayRunner)
+    runner._service_tier = "ultrafast"
+    rk = {"api_key": "k", "base_url": "https://api.openai.com/v1", "provider": "openai", "api_mode": "codex_responses",
+          "command": None, "args": [], "credential_pool": None, "max_tokens": None}
+    assert runner._resolve_turn_agent_config("hi", "gpt-6-astra", rk)["request_overrides"] == {"service_tier": "ultrafast"}
+
+
+def test_cli_refuses_ultrafast_on_a_model_without_it(monkeypatch):
+    import cli as cli_mod
+    from unittest.mock import MagicMock
+
+    stub = SimpleNamespace(service_tier="priority", model="gpt-6-sol", agent=MagicMock(model="gpt-6-sol"),
+                           _fast_command_available=lambda: True)
+    monkeypatch.setattr(cli_mod, "_cprint", lambda *a, **k: None)
+    cli_mod.HermesCLI._handle_fast_command(stub, "/fast ultrafast")
+    assert stub.service_tier == "priority"
+    stub.model = stub.agent.model = "gpt-6-astra"
+    cli_mod.HermesCLI._handle_fast_command(stub, "/fast ultrafast")
+    assert stub.service_tier == "ultrafast"
+
+
+def _usage(prompt_uncached: int, served_tier=None) -> CanonicalUsage:
+    usage = CanonicalUsage(input_tokens=prompt_uncached, output_tokens=10_000, cache_read_tokens=20_000,
+                           cache_write_tokens=5_000)
+    return with_served_service_tier(usage, SimpleNamespace(service_tier=served_tier))
+
+
+@pytest.mark.parametrize("prompt_uncached", [50_000, 400_000])  # below / above the 272K whole-request tier
+def test_served_ultrafast_bills_at_the_ultrafast_row_and_requested_only_does_not(prompt_uncached):
+    for model in _OPENAI_ULTRAFAST_PRICING:
+        standard = estimate_usage_cost(model, _usage(prompt_uncached), provider="openai-api")
+        served_default = estimate_usage_cost(model, _usage(prompt_uncached, "default"), provider="openai-api")
+        ultra = estimate_usage_cost(model, _usage(prompt_uncached, "ultrafast"), provider="openai-api")
+        assert served_default.amount_usd == standard.amount_usd  # asked for Ultrafast, served at Standard
+        assert ultra.pricing_version == _OPENAI_ULTRAFAST_PRICING[model].pricing_version
+        assert ultra.amount_usd == standard.amount_usd * 6  # every bucket, both context tiers
+
+
+def test_served_ultrafast_on_a_model_without_a_published_rate_is_unknown():
+    assert estimate_usage_cost("gpt-6-sol", _usage(1_000, "ultrafast"), provider="openai-api").status == "unknown"
+
+
+def test_codex_stream_assembler_keeps_the_served_tier():
+    from agent.codex_runtime import _consume_codex_event_stream
+
+    done = {"type": "response.completed", "response": {"id": "r1", "status": "completed", "service_tier": "default",
+                                                       "usage": {"input_tokens": 1, "output_tokens": 1}}}
+    events = [{"type": "response.output_text.delta", "delta": "PASS"}, done]
+    assert _consume_codex_event_stream(iter(events), model="gpt-6-astra").service_tier == "default"

tui_gateway/methods_config.py

@@ -233,7 +233,7 @@ def _cfg_get_fast(params):
             else session.get("create_service_tier_override"))
     if tier is None:
         tier = _load_service_tier()
-    return {"value": "fast" if tier == "priority" else "normal"}
+    return {"value": {"priority": "fast", "ultrafast": "ultrafast"}.get(tier, "normal")}
 
 
 def _cfg_get_thinking_mode(params):

tui_gateway/methods_config_set.py

@@ -164,7 +164,7 @@ def _set_model(rid, params, key, value, session):
 
 
 _FAST_WORDS = {"fast": "fast", "on": "fast", "normal": "normal", "off": "normal",
-               "auto": "auto", "cold": "cold"}
+               "auto": "auto", "cold": "cold", "ultrafast": "ultrafast"}
 
 
 def _set_fast(rid, params, key, value, session):
@@ -176,13 +176,14 @@ def _set_fast(rid, params, key, value, session):
         current_tier = session["create_service_tier_override"] or None  # pre-build pin beats global
     else:
         current_tier = _load_service_tier()
+    from agent.fast_mode import STATIC_TIERS, service_tier_word
     if raw == "status":
-        return _kv(rid, key, {"priority": "fast", None: "normal", "": "normal"}.get(current_tier, current_tier))
-    nv = _FAST_WORDS.get(raw, ("normal" if current_tier == "priority" else "fast") if raw in {"", "toggle"} else None)
+        return _kv(rid, key, service_tier_word(current_tier))
+    nv = _FAST_WORDS.get(raw, ("normal" if current_tier in STATIC_TIERS else "fast") if raw in {"", "toggle"} else None)
     if nv is None:
         return _err(rid, 4002, f"unknown fast mode: {value}")
     overrides = None
-    if nv == "fast":
+    if nv in ("fast", "ultrafast"):
         from hermes_cli.models import resolve_fast_mode_overrides
         if agent is not None:
             target_model = getattr(agent, "model", None)
@@ -192,9 +193,10 @@ def _set_fast(rid, params, key, value, session):
         if not target_model:
             return _err(rid, 4002, "fast mode is not available without a selected model")
         overrides = resolve_fast_mode_overrides(target_model, provider=getattr(agent, "provider", None),
-                                                base_url=getattr(agent, "base_url", None))
+                                                base_url=getattr(agent, "base_url", None),
+                                                tier="ultrafast" if nv == "ultrafast" else None)
         if overrides is None:
-            return _err(rid, 4002, "fast mode is not available for this model")
+            return _err(rid, 4002, f"{nv} mode is not available for this model")
     if session is not None:
         # Session-scoped like `reasoning` (global = `--global` / Settings → Model): writing config.yaml
         # here flipped fast mode for every surface. The create override survives rebuilds; "" pins normal.

tui_gateway/methods_slash.py

@@ -323,7 +323,8 @@ def _mirror_prompt(sid, session, agent, arg) -> None:
         agent._cached_system_prompt = None
 
 
-_FAST_TIERS = {"fast": "priority", "on": "priority", "normal": None, "off": None, "auto": "auto", "cold": "cold"}
+_FAST_TIERS = {"fast": "priority", "on": "priority", "normal": None, "off": None, "auto": "auto", "cold": "cold",
+               "ultrafast": "ultrafast"}
 
 
 def _mirror_fast(sid, session, agent, arg) -> None:

tui_gateway/server.py

@@ -29,6 +29,7 @@ from hermes_cli.env_loader import load_hermes_dotenv
 from utils import file_signature, is_truthy_value
 from hermes_state_ids import new_session_id
 from tools.environments.local import hermes_subprocess_env
+from agent.fast_mode import STATIC_TIERS
 from agent.replay_cleanup import canonicalize_replay_history
 from agent.reasoning_effort import clamp_effort, route_supported_efforts
 from agent.compaction_display import project_compaction_message_for_display  # noqa: F401
@@ -1870,12 +1871,10 @@ def _load_reasoning_config(model: str = "") -> dict | None:
     return resolve_reasoning_config(_load_cfg(), model)
 
 
-_SERVICE_TIER_ALIASES = {"fast": "priority", "priority": "priority", "on": "priority", "auto": "auto", "cold": "cold"}
-
-
 def _load_service_tier() -> str | None:
-    raw = str((_load_cfg().get("agent") or {}).get("service_tier", "") or "").strip().lower()
-    return _SERVICE_TIER_ALIASES.get(raw)
+    from agent.fast_mode import parse_service_tier
+
+    return parse_service_tier((_load_cfg().get("agent") or {}).get("service_tier", ""))
 
 
 def _load_provider_routing() -> dict:
@@ -2262,7 +2261,7 @@ def _live_session_identity(session: dict) -> tuple[str, str]:
     return str(model), str(provider or "")
 
 
-def _fast_tier_applies(agent, model: str, provider: str, *, route_known: bool) -> bool:
+def _fast_tier_applies(agent, model: str, provider: str, *, route_known: bool, tier: str | None = None) -> bool:
     """Whether a priority tier reaches this session's route. Every request builder asks the same gate, so a
     profile-wide ``service_tier: fast`` sends nothing to a local server or a proxy, and the session must not
     report Fast there either. ``route_known`` is False while a switch is pending: the agent's base URL still
@@ -2274,7 +2273,7 @@ def _fast_tier_applies(agent, model: str, provider: str, *, route_known: bool) -
             base_url = getattr(agent, "_anthropic_base_url", None)
         base_url = base_url or getattr(agent, "base_url", None)
     try:
-        return resolve_fast_mode_overrides(model, provider=provider or None, base_url=base_url) is not None
+        return resolve_fast_mode_overrides(model, provider=provider or None, base_url=base_url, tier=tier) is not None
     except Exception:
         return False
 
@@ -2325,8 +2324,8 @@ def _session_info(agent, session: dict | None = None) -> dict:
         "provider": pending_provider or provider,
         "reasoning_effort": reasoning_effort, "reasoning_effort_wire": reasoning_effort_wire,
         "service_tier": service_tier,
-        "fast": service_tier == "priority" and _fast_tier_applies(agent, model, pending_provider or provider,
-                                                                  route_known=not pending_provider),
+        "fast": service_tier in STATIC_TIERS and _fast_tier_applies(agent, model, pending_provider or provider,
+                                                                    route_known=not pending_provider, tier=service_tier),
         "yolo": yolo, "approval_mode": approval_mode,
         "tools": dict(mirror.get("tools") or {}) if isinstance(mirror.get("tools"), dict) else {},
         "skills": dict(mirror.get("skills") or {}) if isinstance(mirror.get("skills"), dict) else {},

ui-tui/src/app/slash/commands/session.ts

@@ -26,6 +26,12 @@ const TUI_SESSION_MODEL_RE = new RegExp(`(?:^|\\s)${TUI_SESSION_MODEL_FLAG}(?:\\
 const REASONING_SESSION_FLAGS = new Set(['--session'])
 const REASONING_GLOBAL_FLAGS = new Set(['--global'])
 
+type FastModeWord = 'fast' | 'normal' | 'ultrafast'
+
+// `config.get/set fast` answer fast | ultrafast | normal (auto/cold windows read as normal here).
+const fastModeWord = (value: unknown): FastModeWord =>
+  value === 'fast' || value === 'ultrafast' ? value : 'normal'
+
 const modelValueForConfigSet = (arg: string) => {
   const trimmed = arg.trim()
 
@@ -606,11 +612,11 @@ export const sessionCommands: SlashCommand[] = [
   },
 
   {
-    help: 'toggle fast mode [normal|fast|status|on|off|toggle]',
+    help: 'toggle fast mode [normal|fast|ultrafast|status|on|off|toggle]',
     name: 'fast',
     run: (arg, ctx) => {
       const mode = arg.trim().toLowerCase()
-      const valid = new Set(['', 'status', 'normal', 'fast', 'on', 'off', 'toggle'])
+      const valid = new Set(['', 'status', 'normal', 'fast', 'ultrafast', 'on', 'off', 'toggle'])
 
       if (!valid.has(mode)) {
         return ctx.transcript.sys(t('slashCmd.session.fast.usage'))
@@ -621,7 +627,7 @@ export const sessionCommands: SlashCommand[] = [
           .rpc<ConfigGetValueResponse>('config.get', { key: 'fast', session_id: ctx.sid })
           .then(
             ctx.guarded<ConfigGetValueResponse>(r =>
-              ctx.transcript.sys(t('slashCmd.session.fast.mode', r.value === 'fast' ? 'fast' : 'normal'))
+              ctx.transcript.sys(t('slashCmd.session.fast.mode', fastModeWord(r.value)))
             )
           )
           .catch(ctx.guardedErr)
@@ -631,15 +637,15 @@ export const sessionCommands: SlashCommand[] = [
         .rpc<ConfigSetResponse>('config.set', { key: 'fast', session_id: ctx.sid, value: mode })
         .then(
           ctx.guarded<ConfigSetResponse>(r => {
-            const next = r.value === 'fast' ? 'fast' : 'normal'
+            const next = fastModeWord(r.value)
             ctx.transcript.sys(t('slashCmd.session.fast.mode', next))
             patchUiState(state => ({
               ...state,
               info: state.info
                 ? {
                     ...state.info,
-                    fast: next === 'fast',
-                    service_tier: next === 'fast' ? 'priority' : ''
+                    fast: next !== 'normal',
+                    service_tier: { fast: 'priority', normal: '', ultrafast: 'ultrafast' }[next]
                   }
                 : state.info
             }))

ui-tui/src/i18n/en/slashCmd/session.ts

@@ -30,7 +30,7 @@ export const slashCmdSessionEn = {
     },
     fast: {
       mode: (mode: string) => `fast mode: ${mode}`,
-      usage: 'usage: /fast [normal|fast|status|on|off|toggle]'
+      usage: 'usage: /fast [normal|fast|ultrafast|status|on|off|toggle]'
     },
     indicator: {
       current: (style: string) => `indicator: ${style}`,