返回 CodeWhale
MODEL_PROVIDER_AUDIT.md
根目录 / docs / MODEL_PROVIDER_AUDIT.md
1 # Model & Provider Metadata Audit
2
3 Audit date: **2026-07-12** · Repo state: `main` @ `3e97b278e` (v0.8.68 lane)
4 Scope: every provider and model CodeWhale knows about, their characteristics
5 (context window, max output, reasoning, tools, modalities, pricing, aliases),
6 where each fact lives in code, how the metadata layers interact, and every
7 discrepancy found against the live Models.dev catalog. Intended as the working
8 reference for a future metadata-consolidation pass.
9
10 ---
11
12 ## 1. Executive summary
13
14 1. **The "gpt-5.6-luna 272K" display is correct by design, not a data bug.**
15 On the ChatGPT/Codex OAuth route, CodeWhale shows the context window that
16 OpenAI's account-scoped `/models` endpoint advertises (persisted by the
17 Codex CLI at `~/.codex/models_cache.json`). That cache on this machine
18 (fetched 2026-07-13) advertises `context_window: 272000` for **all**
19 gpt-5.x OAuth models (gpt-5.5, gpt-5.6-sol/terra/luna, gpt-5.4,
20 gpt-5.4-mini; gpt-5.3-codex-spark is 128000). The public **API** route for
21 the same model ids is 1,050,000 (922,000 input + 128,000 output). The
22 deliberate policy — documented in `route_runtime.rs:33-51` and pinned by
23 the test `same_model_id_uses_route_effective_api_and_oauth_metadata`
24 (`model_picker.rs:1660-1706`) — is to never let the OAuth route inherit
25 the API route's bigger window, output cap, or pricing.
26 **Possible UX follow-up (see §8, A1):** label the value as the OAuth-route
27 window (e.g. "272K ctx (ChatGPT route)") so it doesn't read as wrong data.
28
29 2. **Metadata is spread across seven layers** (see §2). The precedence is
30 well-defined and test-guarded, but the *facts* are duplicated in at least
31 four hand-maintained places (`models.rs`, `model_catalog.bundled.json`,
32 `pricing.rs`, `models_dev.bundled.json`), which is where drift creeps in.
33
34 3. **Real drift found** against the live Models.dev catalog — 14 candidate
35 mismatches (§7), the most defensible being GLM window (202,752 vs
36 vendor 200,000), Qwen3.6-27b/35b output caps (we say 262,140; Alibaba says
37 65,536), Grok 4.20 window (we say 2M; xAI catalog row says 1M), and
38 MiniMax-M2 (we say 204,800; catalog says 196,608).
39
40 ---
41
42 ## 2. Metadata architecture — layers and precedence
43
44 Effective precedence for model facts (context/output/reasoning/pricing),
45 lowest → highest, confirmed from `crates/config/src/catalog.rs:9-18`,
46 `CatalogCompiler::compile` (`catalog.rs:582-641`) and
47 `crates/tui/src/provider_lake.rs:54-142`:
48
49 ```
50 (5) legacy static completion lists (DEFAULT_* consts) ← only if catalog has zero rows for provider
51 (4) static code tables crates/tui/src/models.rs ← fallback inside context_window_for_model()
52 (3) bundled Models.dev seed crates/config/assets/models_dev.bundled.json ("NOT a competing source of truth", #4188)
53 + bundled TUI catalog crates/tui/assets/model_catalog.bundled.json (31 entries)
54 (2) live Models.dev catalog https://models.dev/catalog.json → ~/.codewhale/catalog/models-dev-catalog.json (24 h TTL)
55 (1) user / custom overrides (pinned models, custom endpoints, explicit facts)
56 (0) SPECIAL: ChatGPT/Codex OAuth roster ~/.codex/models_cache.json — bypasses the catalog entirely
57 for ApiProvider::OpenaiCodex (provider_lake.rs:131-133, route_runtime.rs:33-51)
58 ```
59
60 Key components:
61
62 | Component | File | Role |
63 |---|---|---|
64 | `ProviderLake` | `crates/tui/src/provider_lake.rs` | Single facade; merges live-over-bundled keyed on `(provider, wire_model_id)`; legacy fallback at `:138-142` |
65 | Models.dev live fetch | `crates/tui/src/models_dev_live.rs` | Background refresh, 24 h TTL, 15 s timeout, atomic disk cache; env knobs `CODEWHALE_MODELS_DEV_URL` / `_PATH` / `CODEWHALE_DISABLE_MODELS_DEV_FETCH` |
66 | Catalog compiler + provenance | `crates/config/src/catalog.rs` | `CatalogSource::{Bundled, Live, UserOverride}`; normalizes Models.dev ids (`moonshotai`→`moonshot`, `togetherai`→`together`, `zhipuai`→`zai`) |
67 | Models.dev schema | `crates/config/src/models_dev.rs` | Network-free deserialization of `{models, providers}` |
68 | Static fact tables | `crates/tui/src/models.rs` | `context_window_for_model` / `max_output_tokens_for_model` / `model_supports_reasoning`; catalog checked first, then explicit `_Nk` suffix hint, then vendor heuristics |
69 | Seeded registry | `crates/tui/src/model_registry.rs` | `ModelMetadata` keyed by id, seeded *from* `models.rs` (drift-guarded by tests); intended future single source |
70 | Pricing | `crates/tui/src/pricing.rs` | Hand-curated USD (+CNY for DeepSeek) rows; catalog USD pricing used when no explicit row |
71 | Codex OAuth roster | `crates/tui/src/codex_model_cache.rs` | Read-only parse of `~/.codex/models_cache.json`; trusted only when fresh (<24 h), else conservative fallback with `context_window: None` (compat floor 128,000: `config/models.rs:126`) |
72 | Agent-crate registry | `crates/agent/src/lib.rs` | 93 `ModelInfo` rows (model×provider), aliases, tools/reasoning flags, resolution fallback chain |
73
74 Fallback heuristics in `models.rs` when nothing above matches:
75 explicit `_Nk` name suffix (8k–1024k) → DeepSeek family (v4 → 1M, legacy →
76 128K) → GPT-5.5/5.6 API → 1.05M → Codex family → 400K → known-model table →
77 any "claude" → 200K → `None` (compaction default threshold 102,400).
78
79 ---
80
81 ## 3. Provider inventory (33 built-in + 1 legacy alias)
82
83 Source: `crates/config/src/provider.rs` (`PROVIDER_REGISTRY`, 33 descriptors),
84 `provider_kind.rs`, `provider_defaults.rs`. Dialect = `WireFormat`
85 (`provider.rs:33`): CC = OpenAI Chat Completions, RESP = OpenAI Responses,
86 AM = Anthropic Messages.
87
88 | id | Display | Default base URL | Default model | Auth (env vars) | Dialect |
89 |---|---|---|---|---|---|
90 | `deepseek` | DeepSeek | `https://api.deepseek.com/beta` | `deepseek-v4-pro` | `DEEPSEEK_API_KEY` | CC |
91 | `deepseek-anthropic` | DeepSeek (Anthropic-compatible) | `https://api.deepseek.com/anthropic` | `deepseek-v4-pro` | `DEEPSEEK_API_KEY` | AM |
92 | `deepseek-cn` *(TUI-only legacy alias)* | DeepSeek (legacy alias) | own config table | — | shares `DEEPSEEK_API_KEY` | CC |
93 | `nvidia-nim` | NVIDIA NIM | `https://integrate.api.nvidia.com/v1` | `deepseek-ai/deepseek-v4-pro` | `NVIDIA_API_KEY`, `NVIDIA_NIM_API_KEY`, `DEEPSEEK_API_KEY` | CC |
94 | `openai` | OpenAI-compatible | `https://api.openai.com/v1` | `deepseek-v4-pro` | `OPENAI_API_KEY` | CC |
95 | `openai-codex` | OpenAI Codex (ChatGPT) | `https://chatgpt.com/backend-api` | `gpt-5.5` | OAuth (`~/.codex/auth.json`) or `OPENAI_CODEX_ACCESS_TOKEN`/`CODEX_ACCESS_TOKEN` | RESP |
96 | `anthropic` | Anthropic | `https://api.anthropic.com` | `claude-sonnet-4-6` | `ANTHROPIC_API_KEY` (no subscription OAuth) | AM |
97 | `atlascloud` | AtlasCloud | `https://api.atlascloud.ai/v1` | `deepseek-ai/deepseek-v4-flash` | `ATLASCLOUD_API_KEY` | CC |
98 | `wanjie-ark` | Wanjie Ark | `https://maas-openapi.wanjiedata.com/api/v1` | `deepseek-reasoner` | `WANJIE_ARK_API_KEY`, `WANJIE_API_KEY`, `WANJIE_MAAS_API_KEY` | CC |
99 | `volcengine` | Volcengine Ark | `https://ark.cn-beijing.volces.com/api/coding/v3` | `DeepSeek-V4-Pro` | `VOLCENGINE_API_KEY`, `VOLCENGINE_ARK_API_KEY`, `ARK_API_KEY` | CC |
100 | `openrouter` | OpenRouter | `https://openrouter.ai/api/v1` | `deepseek/deepseek-v4-pro` | `OPENROUTER_API_KEY` | CC |
101 | `xiaomi-mimo` | Xiaomi MiMo | `https://token-plan-sgp.xiaomimimo.com/v1` (regional cn/sgp/ams + PAYG `api.xiaomimimo.com`) | `mimo-v2.5-pro` | `XIAOMI_MIMO_TOKEN_PLAN_API_KEY`, `MIMO_TOKEN_PLAN_API_KEY`, `XIAOMI_MIMO_API_KEY`, `XIAOMI_API_KEY`, `MIMO_API_KEY` | CC |
102 | `novita` | Novita AI | `https://api.novita.ai/openai/v1` | `deepseek/deepseek-v4-pro` | `NOVITA_API_KEY` | CC |
103 | `fireworks` | Fireworks AI | `https://api.fireworks.ai/inference/v1` | `accounts/fireworks/models/deepseek-v4-pro` | `FIREWORKS_API_KEY` | CC |
104 | `siliconflow` | SiliconFlow | `https://api.siliconflow.com/v1` | `deepseek-ai/DeepSeek-V4-Pro` | `SILICONFLOW_API_KEY` | CC |
105 | `siliconflow-CN` | SiliconFlow (China) | `https://api.siliconflow.cn/v1` | `deepseek-ai/DeepSeek-V4-Pro` | `SILICONFLOW_API_KEY` | CC |
106 | `arcee` | Arcee AI | `https://api.arcee.ai/api/v1` | `trinity-large-thinking` | `ARCEE_API_KEY` | CC |
107 | `moonshot` | Moonshot/Kimi | `https://api.moonshot.ai/v1` (Kimi-for-coding: `https://api.kimi.com/coding/v1`) | `kimi-k2.7-code` | `MOONSHOT_API_KEY`, `KIMI_API_KEY`, or Kimi OAuth | CC |
108 | `sglang` | SGLang (self-hosted) | `http://localhost:30000/v1` | `deepseek-ai/DeepSeek-V4-Pro` | `SGLANG_API_KEY` | CC |
109 | `vllm` | vLLM (self-hosted) | `http://localhost:8000/v1` | `deepseek-ai/DeepSeek-V4-Pro` | `VLLM_API_KEY` | CC |
110 | `ollama` | Ollama (local) | `http://localhost:11434/v1` | `deepseek-v4-flash` | `OLLAMA_API_KEY` | CC |
111 | `huggingface` | Hugging Face | `https://router.huggingface.co/v1` | `deepseek-ai/DeepSeek-V4-Pro` | `HUGGINGFACE_API_KEY`, `HF_TOKEN` | CC |
112 | `together` | Together AI | `https://api.together.xyz/v1` | `deepseek-ai/DeepSeek-V4-Pro` | `TOGETHER_API_KEY` | CC |
113 | `qianfan` | Baidu Qianfan | `https://api.baiduqianfan.ai/v1` | `ernie-4.0-turbo-8k` | `QIANFAN_API_KEY`, `BAIDU_QIANFAN_API_KEY` | CC |
114 | `openmodel` | OpenModel | `https://api.openmodel.ai` | `deepseek-v4-flash` | `OPENMODEL_API_KEY` | AM |
115 | `zai` | Zhipu AI / Z.ai | `https://api.z.ai/api/coding/paas/v4` | `GLM-5.2` | `ZAI_API_KEY`, `Z_AI_API_KEY`, `ZHIPU_API_KEY`, `GLM_API_KEY` | CC |
116 | `stepfun` | StepFun / StepFlash | `https://api.stepfun.ai/v1` | `step-3.7-flash` | `STEPFUN_API_KEY`, `STEP_API_KEY` | CC |
117 | `minimax` | MiniMax | `https://api.minimax.io/v1` | `MiniMax-M3` | `MINIMAX_API_KEY` | CC |
118 | `deepinfra` | DeepInfra | `https://api.deepinfra.com/v1/openai` | `deepseek-ai/DeepSeek-V4-Pro` | `DEEPINFRA_API_KEY`, `DEEPINFRA_TOKEN` | CC |
119 | `sakana` | Sakana AI (Fugu) | `https://api.sakana.ai/v1` | `fugu` | `FUGU_API_KEY`, `SAKANA_API_KEY` | CC |
120 | `longcat` | Meituan LongCat | `https://api.longcat.chat/openai/v1` | `LongCat-2.0` | `LONGCAT_API_KEY` | CC |
121 | `meta` | Meta Model API | `https://api.meta.ai/v1` | `muse-spark-1.1` | `META_MODEL_API_KEY`, `MODEL_API_KEY` | CC |
122 | `xai` | xAI | `https://api.x.ai/v1` | `grok-4.5` | `XAI_API_KEY` or Grok OAuth (`~/.grok/auth.json`) | CC |
123 | `custom` | Custom (OpenAI-compatible) | per `[providers.<name>]` table | per table | per-entry `api_key_env` | CC |
124
125 OAuth routes (token precedence: route OAuth → CLI key → provider/root config
126 → ambient env; `crates/tui/src/config.rs:3494-3566`):
127
128 - **OpenAI Codex/ChatGPT** — Codex CLI login (`~/.codex/auth.json`);
129 account-scoped model roster from `~/.codex/models_cache.json`; usage is
130 subscription-scoped so **no dollar pricing is shown** on this route
131 (`pricing.rs:127-129, 334-338`).
132 - **xAI Grok** — `[providers.xai] auth_mode = "oauth"`, reuses `~/.grok/auth.json`
133 or device-code login.
134 - **Moonshot/Kimi** — Kimi CLI OAuth for the coding endpoint.
135 - **No Anthropic/Claude subscription OAuth exists** — API key only.
136
137 ---
138
139 ## 4. First-class model metadata (the curated set)
140
141 Merged view of the four fact sources for the models CodeWhale makes explicit
142 promises about. Columns: **Ctx** = context window (tokens), **Out** = max
143 output, **R** = emits reasoning, pricing = USD per 1M tokens as
144 **cache-hit / input / output** from `pricing.rs` (catalog rows have no
145 cache-hit discount → hit = input).
146
147 ### 4.1 DeepSeek (first-class)
148
149 | Model | Ctx | Out | R | Pricing (hit/in/out USD) | Notes |
150 |---|---|---|---|---|---|
151 | `deepseek-v4-pro` | 1,000,000 | 384,000 | ✓ | 0.003625 / 0.435 / 0.87 (+CNY 0.025/3/6) | post-2026-05-31 adjusted rate is permanent (#2489) |
152 | `deepseek-v4-flash` | 1,000,000 | 384,000 | ✓ | 0.0028 / 0.14 / 0.28 (+CNY 0.02/1/2) | |
153 | `deepseek-reasoner`, legacy v3.x, `deepseek-coder*` | 128,000 | — | v4-only | flash rates for non-pro | legacy fallback window (`models.rs:7`) |
154 | `deepseek-ai/*` (NIM-hosted) | 1,000,000 | 384,000 | ✓ | **intentionally unpriced** | NVIDIA terms ≠ DeepSeek platform pricing (`pricing.rs:133-137`) |
155 | any id with `-Nk` suffix | N×1000 | — | — | — | vendor-agnostic served-name hint, 8k–1024k |
156
157 ### 4.2 OpenAI
158
159 | Model | Ctx (API route) | Ctx (OAuth route) | Out | R | Pricing (hit/in/out) | Notes |
160 |---|---|---|---|---|---|---|
161 | `gpt-5.6` (alias → sol) | 1,050,000 | 272,000¹ | 128,000 | ✓ | 0.50 / 5.00 / 30.00 | Models.dev: input limit 922,000 |
162 | `gpt-5.6-sol` | 1,050,000 | 272,000¹ | 128,000 | ✓ | 0.50 / 5.00 / 30.00 | efforts low→ultra on OAuth |
163 | `gpt-5.6-terra` | 1,050,000 | 272,000¹ | 128,000 | ✓ | 0.25 / 2.50 / 15.00 | |
164 | `gpt-5.6-luna` | 1,050,000 | 272,000¹ | 128,000 | ✓ | 0.10 / 1.00 / 6.00 | cost-efficient tier; **the "272k looks wrong" report — see §6** |
165 | `gpt-5.5` | 1,050,000 | 272,000¹ | 128,000 | ✓ | 0.50 / 5.00 / 30.00 | date snapshots (`gpt-5.5-YYYY-MM-DD`) too |
166 | `gpt-5.5-pro` | 1,050,000 | n/a | 128,000 | ✓ | 30.00 / 30.00 / 180.00 | no cached-input discount |
167 | `gpt-5-codex` | 400,000 | per roster | 128,000 | ✓ | 0.125 / 1.25 / 10.00 | deprecated upstream on OAuth path |
168 | `gpt-5.3-codex` | 400,000 | per roster | 128,000 | ✓ | 0.175 / 1.75 / 14.00 | |
169 | other codex ids (`gpt-5.1-codex[-mini/-max]`, `gpt-5.2-codex`, `codex-gpt-5.5`, `chatgpt-gpt-5.5`, `gpt-5.5-codex[-preview]`…) | 400,000 | per roster | 128,000 | ✓ | unpriced | recognized by `is_openai_codex_model` (`models.rs:491-507`) |
170 | `gpt-5.5-nano` | *unknown* | *unknown* | — | ✗ | — | deliberately unrecognized (`models.rs:788-790`) |
171
172 ¹ OAuth window is whatever `~/.codex/models_cache.json` advertises for the
173 account; on this machine (2026-07-13) it is 272,000 for all gpt-5.x list
174 models and 128,000 for `gpt-5.3-codex-spark`. Stale/missing cache → no ctx
175 shown in picker; runtime compat floor 128,000. OAuth route never shows pricing.
176
177 ### 4.3 Anthropic
178
179 | Model | Ctx | Out | R | Pricing (hit/in/out) | Notes |
180 |---|---|---|---|---|---|
181 | `claude-opus-4-8` | 1,000,000 | 128,000 | ✓ | 0.50 / 5.00 / 25.00 | |
182 | `claude-sonnet-4-6` | 1,000,000 | 128,000 | ✓ | 0.30 / 3.00 / 15.00 | out raised 64K→128K (2026-07-09 audit) |
183 | `claude-sonnet-5` | 1,000,000 | 128,000 | ✓ | intro 0.20/2.00/10.00 until 2026-08-31, then 0.30/3.00/15.00 | time-aware in `pricing.rs:263-273` |
184 | `claude-fable-5` | 1,000,000 | 128,000 | ✓ | 1.00 / 10.00 / 50.00 | tokenizer yields ~30% more tokens — raw rate comparisons undercount cost (`pricing.rs:178-182`) |
185 | `claude-haiku-4-5` | 200,000 | 64,000 | ✗ | 0.10 / 1.00 / 5.00 | |
186 | any other `claude*` | 200,000 | — | ✗ | — | family fallback |
187
188 ### 4.4 Moonshot / Kimi
189
190 | Model | Ctx | Out | R | Pricing | Notes |
191 |---|---|---|---|---|---|
192 | `kimi-k2.7-code` (± `moonshotai/`) | 262,144 | 262,144 | ✓ | 0.19 / 0.95 / 4.00 | |
193 | `kimi-k2.6` (± prefix, `:free`) | 262,144 | 262,144 | ✓ | 0.16 / 0.95 / 4.00 | |
194 | `kimi-for-coding` | 262,144 | 262,144 | ✓ | — | stable coding route; rides K2.7 path; **not in agent-crate registry or catalogs** (§7 D-12) |
195 | any bare `kimi-*` | — | — | ✓ | — | prefix rule: always reasoning (#3016) |
196
197 ### 4.5 Z.ai / GLM
198
199 | Model | Ctx | Out | R | Pricing | Notes |
200 |---|---|---|---|---|---|
201 | `glm-5.2` (± `z-ai/`) | 1,000,000 | 131,072 | ✓ | 0.26 / 1.40 / 4.40 | |
202 | `glm-5.3` (± `z-ai/`) | 1,000,000 | 131,072 | ✓ | | **not released by Z.ai as of 2026-08-03** — registered id only; limits/reasoning inherited wholesale from `glm-5.2` pending official Z.ai release metadata, not measured; no published price |
203 | `glm-5.1` (± `z-ai/`) | 202,752 | 131,072 | ✓ | 0.26 / 1.40 / 4.40 | vendor page says 200K (§7 D-4) |
204 | `glm-5-turbo` (± `z-ai/`) | 202,752 | 131,072 | ✓ | 0.24 / 1.20 / 4.00 | fast **text** sibling |
205 | `glm-5v-turbo` (± `z-ai/`) | 202,752 | — | ✗ | — | **vision** model, distinct from 5-turbo |
206
207 ### 4.6 MiniMax
208
209 | Model | Ctx | Out | R | Pricing | Notes |
210 |---|---|---|---|---|---|
211 | `minimax-m3` (± `minimax/`, `MiniMax-M3`) | 1,000,000 | 524,288 | ✓ | 0.06 / 0.30 / 1.20 | catalog says out 128,000 (§7 D-5) |
212 | `minimax-m2.7` (± prefix, `-highspeed`) | 204,800 | 131,072 (catalog) | ✓ | 0.3 / 0.3 / 1.2 (catalog) | |
213 | `minimax-m2.5` / `m2.1` (± `-highspeed`) | 204,800 | — | ✓ | — | |
214 | `minimax-m2` | 204,800 | — | ✓ | — | catalog says 196,608 (§7 D-6) |
215
216 ### 4.7 Qwen (OpenRouter-routed)
217
218 | Model | Ctx | Out | R | Pricing | Notes |
219 |---|---|---|---|---|---|
220 | `qwen/qwen3.6-flash` | 1,000,000 | 65,536 | ✓ | 0.1875 / 0.1875 / 1.125 | |
221 | `qwen/qwen3.6-plus` | 1,000,000 | 65,536 | ✓ | 0.325 / 0.325 / 1.95 | |
222 | `qwen/qwen3.6-35b-a3b` | 262,144 | 262,140 | ✓ | 0.05 / 0.14 / 1.00 | out cap suspect (§7 D-7) |
223 | `qwen/qwen3.6-27b` | 262,144 | 262,140 | ✓ | 0.15 / 0.285 / 2.40 | out cap suspect (§7 D-7) |
224 | `qwen/qwen3.6-max-preview` | 262,144 | 65,536 | ✓ | 1.04 / 1.04 / 6.24 | catalog ctx 245,800 (§7 D-8) |
225 | `qwen/qwen3.7-plus` | — | — | — | 0.064 / 0.32 / 1.28 | **priced but no ctx/out/reasoning rows** (§7 D-13) |
226 | `qwen/qwen3.7-max` | — | — | — | 0.25 / 1.25 / 3.75 | in agent registry; same gap (§7 D-13) |
227
228 ### 4.8 Xiaomi MiMo
229
230 | Model | Ctx | Out | R | Pricing | Notes |
231 |---|---|---|---|---|---|
232 | `mimo-v2.5-pro` (± `xiaomi/`, `-ultraspeed`) | 1,000,000 | 131,072 | ✓ | intentionally unknown | Token-Plan credit billing, no balance endpoint |
233 | `mimo-v2.5` (± `xiaomi/`) | 1,000,000 | 131,072 | ✓ | unknown | omni (text+image) |
234 | `mimo-v2.5-asr` | 8,000 | 2,048 | ✗ | unknown | speech-to-text |
235 | `mimo-v2.5-tts[-voicedesign/-voiceclone]`, `mimo-v2-tts` | 8,000 | 8,192 | ✗ | unknown | TTS family |
236
237 ### 4.9 xAI / Grok
238
239 | Model | Ctx | Out | R | Pricing | Notes |
240 |---|---|---|---|---|---|
241 | `grok-4.5` | 500,000 | — | ✓ | — | |
242 | `grok-4.3` | 1,000,000 | — | ✓ | — | |
243 | `grok-build` | 512,000 | — | ✓ | — | not in Models.dev |
244 | `grok-composer-2.5-fast` | 200,000 | — | ✗ | — | not in Models.dev |
245 | `grok-4.20-0309-reasoning` / `-non-reasoning` | 2,000,000 | — | ✓/✗ | — | catalog row says 1M (§7 D-9); **no Grok pricing rows at all** (§7 D-14) |
246
247 ### 4.10 Others
248
249 | Model | Provider | Ctx | Out | R | Pricing | Notes |
250 |---|---|---|---|---|---|---|
251 | `trinity-large-thinking` (± `arcee-ai/`) | Arcee | 262,144 | 262,144 | ✓ | 0.25 / 0.25 / 0.80 | out=ctx suspect (§7 D-2) |
252 | `trinity-large-preview` | Arcee | 262,144 | — | ✗ | — | catalog row says 131,000 (§7 D-3) |
253 | `trinity-mini` | Arcee | 128,000 | 64,000 (bundled) | ✗ | 0.045 / 0.045 / 0.15 | catalog says ctx 131,072 (§7 D-1) |
254 | `step-3.7-flash` | StepFun | 256,000 | 256,000 | ✗ | 0.2 / 0.2 / 1.15 | third-party sourced (models.dev + AA) |
255 | `fugu` | Sakana | — | — | ✗ | — | in agent registry only |
256 | `fugu-ultra` / `fugu-ultra-20260615` | Sakana | 1,000,000 | 131,000 | ✓ | 5.0 / 5.0 / 30.0 | limits third-party sourced (Requesty); Sakana's own >272K price tier confirms ctx > 272K |
257 | `muse-spark-1.1` | Meta | 1,000,000 | 32,000 | ✓ | 1.25 / 1.25 / 4.25 | |
258 | `LongCat-2.0` | LongCat | — | — | ✓ | — | agent registry only; no fact rows (§7 D-15) |
259 | `tencent/hy3-preview` | OpenRouter | 262,144 | — | ✓ | 0.021 / 0.063 / 0.21 | catalog says 256,000 (§7 D-10) |
260 | `google/gemma-4-31b-it` (± `:free`) | OpenRouter | 262,144 | 16,384 (paid) / 32,768 (free) | ✓ | 0.09 / 0.12 / 0.35 | out cap vs catalog 131,072 (§7 D-11) |
261 | `google/gemma-4-26b-a4b-it` (± `:free`) | OpenRouter | 262,144 | 32,768 (free) | ✓ | 0.06 / 0.06 / 0.33 | |
262 | `nvidia/nemotron-3-ultra-550b-a55b` (± `:free`) | OpenRouter | 1,000,000 | 16,384 (paid) / 65,536 (free) | ✓ | 0.10 / 0.50 / 2.20 | catalog out 65,000 (§7 D-11) |
263 | `nvidia/nemotron-3-nano-omni-30b-a3b-reasoning:free` | OpenRouter | 262,144 | 65,536 | ✓ | — | |
264 | `ernie-4.0-turbo-8k` | Qianfan | 8,000 (via `-8k` hint) | — | ✗ | — | Qianfan default; no explicit rows |
265
266 ---
267
268 ## 5. Bundled TUI catalog (`crates/tui/assets/model_catalog.bundled.json`)
269
270 31 entries, `fetched_at: 2026-07-06`, effectively-infinite TTL. Adds
271 modalities and (for some rows) USD pricing on top of §4 facts. Highlights:
272
273 - Modalities `text,image`: gpt-5.6 family, muse-spark-1.1, minimax-m3, mimo-v2.5.
274 - `text,image,audio`: mimo-v2.5-pro (the only audio-input row).
275 - `gpt-5.6` carries `provider_model_id: gpt-5.6-sol` (alias resolution).
276 - Pricing present only for: codex models, kimi-k2.7-code, glm-5.2,
277 minimax-m2.7, trinity-mini, claude family, step-3.7-flash, fugu-ultra.
278 - The bundled `models_dev.bundled.json` (config crate) is smaller: 2 models
279 (deepseek-v4-pro/flash), 14 providers, 42 chat offerings; demoted to
280 offline-fallback-only by #4188.
281
282 ## 6. The gpt-5.6-luna 272K finding (root cause, resolved)
283
284 - **What you saw:** model picker hint "272K ctx" for gpt-5.6-luna.
285 - **Where it comes from:** `~/.codex/models_cache.json` — the ChatGPT/Codex
286 OAuth `/models` roster persisted by the Codex CLI. On this machine every
287 gpt-5.x OAuth entry advertises `context_window: 272000`.
288 - **Why CodeWhale shows it:** for `ApiProvider::OpenaiCodex` the picker and
289 runtime use the OAuth-advertised window **exclusively** — never the API
290 route's 1,050,000 — because the OAuth offering genuinely has the smaller
291 window and different (subscription) billing. Precedence code:
292 `model_picker.rs:1105-1111`; runtime: `route_runtime.rs:33-51`; compaction
293 consumes it via `effective_context_window` (`compaction.rs:98-104`).
294 Pinned by tests (`model_picker.rs:1660-1706`,
295 `commands/groups/debug/tests.rs:96-111`).
296 - **Is 272,000 plausible?** Yes: 400,000 (codex-class window) − 128,000
297 (output) = 272,000 usable input; OpenAI advertises the input budget as the
298 OAuth "context_window". The same-name API models are 1,050,000 total /
299 922,000 input.
300 - **Verdict:** data is correct per route; the *presentation* invites the
301 "that looks wrong" reaction. See action A1.
302
303 ## 7. Discrepancies & gaps (vs live Models.dev catalog, fetched 2026-07-12)
304
305 Confidence key: **vendor** = the live row is the vendor's own provider entry
306 (strong signal); **aggregator** = row from a reseller (Vercel/OpenRouter etc.,
307 weaker — verify against the vendor's docs before changing anything).
308
309 | # | Model | CodeWhale says | Live catalog says | Source | Assessment |
310 |---|---|---|---|---|---|
311 | D-1 | `trinity-mini` | ctx 128,000 / out 64,000 (bundled) | 131,072 / 131,072 | aggregator | verify vs Arcee docs |
312 | D-2 | `trinity-large-thinking` | out 262,144 (= full ctx — suspicious) | ctx 262,100 / out 80,000 | aggregator | out=ctx is a common data-entry smell; verify |
313 | D-3 | `trinity-large-preview` | ctx 262,144 | 131,000 / 131,000 | aggregator | verify |
314 | D-4 | `glm-5.1`, `glm-5-turbo`, `glm-5v-turbo` | ctx 202,752 | 200,000 | **vendor (zai)** | 202,752 = 198×1024; likely fine (marketing 200K vs binary), document choice |
315 | D-5 | `minimax-m3` | out 524,288 | out 128,000 | vendor (coding-plan row) | route-dependent; verify per MiniMax platform docs |
316 | D-6 | `minimax-m2` | ctx 204,800 | 196,608 | vendor (coding-plan row) | verify |
317 | D-7 | `qwen3.6-35b-a3b`, `qwen3.6-27b` | out 262,140 | out 65,536 | **vendor (alibaba)** | our 262,140 looks wrong (≈ctx); likely fix to 65,536 |
318 | D-8 | `qwen3.6-max-preview` | ctx 262,144 | 245,800 | vendor (alibaba-cn) | verify |
319 | D-9 | `grok-4.20-0309-*` | ctx 2,000,000 | 1,000,000 | vendor (xai) | 2M matches xAI's fast-endpoint marketing; catalog row may be the standard endpoint — verify |
320 | D-10 | `tencent/hy3-preview` | ctx 262,144 | 256,000 | aggregator | minor; verify |
321 | D-11 | `gemma-4-31b-it` out 16,384; `nemotron-3-ultra` out 16,384 | 131,072 / 65,000 | aggregator | per-host caps differ; ours were OpenRouter-specific — document as route-scoped |
322 | D-12 | `kimi-for-coding` | in `models.rs` only | absent from live catalog, agent registry, bundled catalogs | — | add to agent registry + bundled catalog or document as models.rs-only route |
323 | D-13 | `qwen/qwen3.7-plus`, `qwen/qwen3.7-max` | priced (and 3.7-max in agent registry) but **no ctx/out/reasoning rows** | present upstream | — | add fact rows |
324 | D-14 | all Grok models | **no pricing rows** | xai: e.g. grok-4.5 $2/$6, grok-4.3 $1.25/$2.5 | vendor | add pricing (catalog passthrough may already cover once live rows resolve — verify `resolved_usd_pricing` path) |
325 | D-15 | `LongCat-2.0`, `fugu` (base), `ernie-4.0-turbo-8k` | agent-registry/default only; no metadata rows | — | — | add facts or mark best-effort |
326 | D-16 | `mimo-v2.5*` | ctx 1,000,000 | 1,048,576 | vendor (token-plan) | cosmetic (1M vs 2^20); document choice |
327 | D-17 | `fugu-ultra` | out 131,000 | sakana row: out 1,000,000 | vendor row suspect (out=ctx) | keep ours (Requesty-sourced) unless Sakana docs say otherwise |
328
329 Structural observations:
330
331 - **Four hand-maintained fact stores** must currently be updated in lockstep:
332 `models.rs`, `model_catalog.bundled.json`, `pricing.rs`,
333 `models_dev.bundled.json`. `model_registry.rs` is the intended chokepoint
334 but is not yet consumed by production call sites (its module docs say so).
335 - The agent crate's `ModelRegistry` (93 rows) duplicates provider/alias data
336 with its own `supports_tools`/`supports_reasoning` flags — a fifth store.
337 - `CurrencyPricing` has no **cache-write** field, so Anthropic (1.25–2× input)
338 and Qwen 3.7 cache-write rates are silently dropped from cost estimates.
339 - The bundled TUI catalog cannot carry a cache-read rate, forcing Anthropic
340 rows to live in `pricing.rs` above the catalog.
341
342 ## 8. Action items (for the implementation pass)
343
344 | ID | Action | Where |
345 |---|---|---|
346 | A1 | Label OAuth-route context in the picker (e.g. "272K ctx · ChatGPT route") so account-scoped windows don't read as wrong data; optionally show the API-route window alongside | `model_picker.rs:1187-1192` hint renderer |
347 | A2 | Fix Qwen3.6-27b/35b output caps (262,140 → 65,536, pending vendor-doc check) | `models.rs:371`, tests |
348 | A3 | Verify & reconcile D-1/2/3 (Arcee), D-5/6 (MiniMax), D-8 (Qwen max-preview), D-9 (Grok 4.20), D-10 (HY3) against vendor docs | `models.rs`, bundled catalogs |
349 | A4 | Add fact rows for `qwen3.7-plus/max`, `LongCat-2.0`, Grok pricing | `models.rs`, `pricing.rs` |
350 | A5 | Register `kimi-for-coding` in the agent-crate registry + bundled catalog | `crates/agent/src/lib.rs`, assets |
351 | A6 | Add cache-write field to `CurrencyPricing` and to the catalog schema; move Anthropic rows into the catalog once it can carry cache rates | `pricing.rs:97-109`, `catalog.rs` |
352 | A7 | Complete the #3071/#3073 migration: make production call sites consume `model_registry::lookup` so the fact stores collapse to one | `model_registry.rs` |
353 | A8 | Consider folding the agent-crate `ModelRegistry` flags into the same chokepoint (or generating them from it) | `crates/agent/src/lib.rs` |
354 | A9 | Refresh `model_catalog.bundled.json` + `models_dev.bundled.json` from the 2026-07-12 live snapshot as part of each release lane | assets |
355
356 ## Appendix A — agent-crate registry (model × provider, 93 rows)
357
358 Source: `crates/agent/src/lib.rs` `ModelRegistry::default()`. Flags:
359 T = supports tools, R = supports reasoning. Aliases are case-insensitive.
360
361 | Wire model id | Provider | T | R | Aliases |
362 |---|---|---|---|---|
363 | `deepseek-v4-pro` | Deepseek | ✓ | ✓ | — |
364 | `deepseek-v4-flash` | Deepseek | ✓ | ✓ | deepseek-chat, deepseek-reasoner, deepseek-r1, deepseek-v3, deepseek-v3.2 |
365 | `deepseek-ai/deepseek-v4-pro` | NvidiaNim | ✓ | ✓ | deepseek-v4-pro, nvidia-deepseek-v4-pro, nim-deepseek-v4-pro |
366 | `deepseek-ai/deepseek-v4-flash` | NvidiaNim | ✓ | ✓ | deepseek-v4-flash, deepseek-chat, deepseek-reasoner, nvidia-deepseek-v4-flash, nim-deepseek-v4-flash |
367 | `deepseek-v4-pro` | Openai | ✓ | ✓ | openai-compatible-deepseek-v4-pro |
368 | `deepseek-v4-flash` | Openai | ✓ | ✓ | openai-compatible-deepseek-v4-flash |
369 | `gpt-5.6` | Openai | ✓ | ✓ | gpt56 |
370 | `gpt-5.6-sol` | Openai | ✓ | ✓ | gpt56-sol |
371 | `gpt-5.6-terra` | Openai | ✓ | ✓ | gpt56-terra |
372 | `gpt-5.6-luna` | Openai | ✓ | ✓ | gpt56-luna |
373 | `deepseek-ai/deepseek-v4-flash` | Atlascloud | ✓ | ✓ | deepseek-v4-flash, atlascloud-deepseek-v4-flash |
374 | `deepseek-ai/deepseek-v4-pro` | Atlascloud | ✓ | ✓ | deepseek-v4-pro, atlascloud-deepseek-v4-pro |
375 | `deepseek-reasoner` | WanjieArk | ✓ | ✓ | wanjie-deepseek-reasoner, ark-wanjie-deepseek-reasoner |
376 | `DeepSeek-V4-Pro` | Volcengine | ✓ | ✓ | deepseek-v4-pro, volcengine-deepseek-v4-pro, ark-deepseek-v4-pro |
377 | `DeepSeek-V4-Flash` | Volcengine | ✓ | ✓ | deepseek-v4-flash, deepseek-chat, volcengine-deepseek-v4-flash, ark-deepseek-v4-flash |
378 | `trinity-large-thinking` | Arcee | ✓ | ✓ | trinity, arcee-trinity, arcee-trinity-large-thinking |
379 | `trinity-large-preview` | Arcee | ✓ | ✗ | arcee-trinity-large-preview |
380 | `deepseek/deepseek-v4-pro` | Openrouter | ✓ | ✓ | deepseek-v4-pro, openrouter-deepseek-v4-pro |
381 | `deepseek/deepseek-v4-flash` | Openrouter | ✓ | ✓ | deepseek-v4-flash, deepseek-chat, deepseek-reasoner, openrouter-deepseek-v4-flash |
382 | `arcee-ai/trinity-large-thinking` | Openrouter | ✓ | ✓ | trinity, trinity-large-thinking, arcee-trinity-large-thinking |
383 | `xiaomi/mimo-v2.5-pro` | Openrouter | ✓ | ✓ | openrouter-mimo-v2.5-pro, openrouter-xiaomi-mimo-v2.5-pro |
384 | `xiaomi/mimo-v2.5` | Openrouter | ✓ | ✓ | openrouter-mimo-v2.5, openrouter-xiaomi-mimo-v2.5 |
385 | `qwen/qwen3.6-flash` | Openrouter | ✓ | ✓ | qwen3.6-flash, qwen-3.6-flash |
386 | `qwen/qwen3.6-35b-a3b` | Openrouter | ✓ | ✓ | qwen3.6-35b-a3b, qwen-3.6-35b-a3b |
387 | `qwen/qwen3.6-max-preview` | Openrouter | ✓ | ✓ | qwen3.6-max-preview, qwen-3.6-max-preview, qwen-max-preview |
388 | `qwen/qwen3.6-27b` | Openrouter | ✓ | ✓ | qwen3.6-27b, qwen-3.6-27b |
389 | `qwen/qwen3.6-plus` | Openrouter | ✓ | ✓ | qwen3.6-plus, qwen-3.6-plus |
390 | `qwen/qwen3.7-max` | Openrouter | ✓ | ✓ | qwen3.7-max, qwen-3.7-max |
391 | `moonshotai/kimi-k2.7-code` | Openrouter | ✓ | ✓ | kimi-k2.7-code, openrouter-kimi-k2.7-code |
392 | `moonshotai/kimi-k2.6` | Openrouter | ✓ | ✓ | openrouter-kimi-k2.6 |
393 | `minimax/minimax-m3` | Openrouter | ✓ | ✓ | minimax-m3, minimax-m-3, openrouter-minimax-m3 |
394 | `minimax/minimax-m2.7` | Openrouter | ✓ | ✓ | minimax-2.7, minimax-2-7, openrouter-minimax-2.7 |
395 | `z-ai/glm-5.1` | Openrouter | ✓ | ✓ | glm-5.1, zai-glm-5.1 |
396 | `z-ai/glm-5.2` | Openrouter | ✓ | ✓ | glm-5.2, zai-glm-5.2 |
397 | `z-ai/glm-5.3` | Openrouter | ✓ | ✓ | glm-5.3, zai-glm-5.3 |
398 | `z-ai/glm-5-turbo` | Openrouter | ✓ | ✓ | glm-5-turbo, zai-glm-5-turbo |
399 | `tencent/hy3-preview` | Openrouter | ✓ | ✓ | hy3-preview, tencent-hy3-preview |
400 | `google/gemma-4-31b-it` | Openrouter | ✓ | ✓ | gemma-4-31b, gemma-4-31b-it |
401 | `google/gemma-4-26b-a4b-it` | Openrouter | ✓ | ✓ | gemma-4-26b-a4b, gemma-4-26b-a4b-it |
402 | `nvidia/nemotron-3-nano-omni-30b-a3b-reasoning:free` | Openrouter | ✓ | ✓ | nemotron-3-nano-omni, nemotron-3-nano-omni-reasoning |
403 | `nvidia/nemotron-3-ultra-550b-a55b` | Openrouter | ✓ | ✓ | nvidia/nemotron-3-ultra, nemotron-3-ultra, nemotron-3-ultra-550b-a55b, nvidia-nemotron-3-ultra, nvidia-nemotron-3-ultra-550b-a55b |
404 | `GLM-5.2` | Zai | ✓ | ✓ | glm-5.2, glm-5-2, zai-glm-5.2, zai-glm-5-2 |
405 | `GLM-5.3` | Zai | ✓ | ✓ | glm-5.3, glm-5-3, zai-glm-5.3, zai-glm-5-3 |
406 | `GLM-5.1` | Zai | ✓ | ✓ | glm-5.1, glm-5-1, zai-glm-5.1, zai-glm-5-1 |
407 | `GLM-5-Turbo` | Zai | ✓ | ✓ | glm-5-turbo, glm-5turbo, zai-glm-5-turbo |
408 | `mimo-v2.5-pro` | XiaomiMimo | ✓ | ✓ | mimo, pro, xiaomi-mimo-v2.5-pro, xiaomi-mimo-v2-5-pro |
409 | `mimo-v2.5` | XiaomiMimo | ✓ | ✓ | omni, mimo-omni, v2.5-omni, mimo-v2.5-omni, xiaomi-mimo-v2.5, xiaomi-mimo-v2.5-omni |
410 | `mimo-v2.5-asr` | XiaomiMimo | ✗ | ✗ | asr, speech-to-text, transcribe |
411 | `mimo-v2.5-tts` | XiaomiMimo | ✗ | ✗ | tts, speech, mimo-tts |
412 | `mimo-v2.5-tts-voicedesign` | XiaomiMimo | ✗ | ✗ | voicedesign, voice-design, mimo-voice-design |
413 | `mimo-v2.5-tts-voiceclone` | XiaomiMimo | ✗ | ✗ | voiceclone, voice-clone, mimo-voice-clone |
414 | `mimo-v2-tts` | XiaomiMimo | ✗ | ✗ | mimo-v2-speech |
415 | `deepseek/deepseek-v4-pro` | Novita | ✓ | ✓ | deepseek-v4-pro, novita-deepseek-v4-pro |
416 | `deepseek/deepseek-v4-flash` | Novita | ✓ | ✓ | deepseek-v4-flash, deepseek-chat, deepseek-reasoner, novita-deepseek-v4-flash |
417 | `accounts/fireworks/models/deepseek-v4-pro` | Fireworks | ✓ | ✓ | deepseek-v4-pro, fireworks-deepseek-v4-pro |
418 | `deepseek-ai/DeepSeek-V4-Pro` | Siliconflow | ✓ | ✓ | deepseek-v4-pro, deepseek-reasoner, deepseek-r1, siliconflow-deepseek-v4-pro |
419 | `deepseek-ai/DeepSeek-V4-Flash` | Siliconflow | ✓ | ✓ | deepseek-v4-flash, deepseek-chat, deepseek-v3, siliconflow-deepseek-v4-flash |
420 | `kimi-k2.7-code` | Moonshot | ✓ | ✓ | kimi, kimi-k2, kimi-k2.7, kimi-code, moonshot-kimi-k2.7-code |
421 | `kimi-k2.6` | Moonshot | ✓ | ✓ | moonshot-kimi-k2.6 |
422 | `deepseek-ai/DeepSeek-V4-Pro` | Sglang | ✓ | ✓ | deepseek-v4-pro, sglang-deepseek-v4-pro |
423 | `deepseek-ai/DeepSeek-V4-Flash` | Sglang | ✓ | ✓ | deepseek-v4-flash, deepseek-chat, deepseek-reasoner, sglang-deepseek-v4-flash |
424 | `deepseek-ai/DeepSeek-V4-Pro` | Vllm | ✓ | ✓ | deepseek-v4-pro, vllm-deepseek-v4-pro |
425 | `deepseek-ai/DeepSeek-V4-Flash` | Vllm | ✓ | ✓ | deepseek-v4-flash, deepseek-chat, deepseek-reasoner, vllm-deepseek-v4-flash |
426 | `deepseek-v4-flash` | Ollama | ✓ | ✓ | — (Ollama also accepts any name as-is) |
427 | `deepseek-ai/DeepSeek-V4-Pro` | Huggingface | ✓ | ✓ | deepseek-v4-pro, hf-deepseek-v4-pro |
428 | `deepseek-ai/DeepSeek-V4-Flash` | Huggingface | ✓ | ✓ | deepseek-v4-flash, deepseek-chat, deepseek-reasoner, hf-deepseek-v4-flash |
429 | `deepseek-ai/DeepSeek-V4-Pro` | Together | ✓ | ✓ | deepseek-v4-pro, together-deepseek-v4-pro |
430 | `deepseek-ai/DeepSeek-V4-Flash` | Together | ✓ | ✓ | deepseek-v4-flash, deepseek-chat, together-deepseek-v4-flash |
431 | `gpt-5.5` | OpenaiCodex | ✓ | ✓ | codex-gpt-5.5, chatgpt-gpt-5.5 |
432 | `claude-opus-4-8` | Anthropic | ✓ | ✓ | opus, claude-opus |
433 | `claude-sonnet-4-6` | Anthropic | ✓ | ✓ | sonnet, claude-sonnet |
434 | `claude-haiku-4-5` | Anthropic | ✓ | ✗ | haiku, claude-haiku |
435 | `deepseek-v4-flash` | Openmodel | ✓ | ✓ | openmodel, openmodel-deepseek |
436 | `step-3.7-flash` | Stepfun | ✓ | ✗ | stepfun, stepflash |
437 | `MiniMax-M3` | Minimax | ✓ | ✓ | minimax, minimax-m3, minimax-m-3 |
438 | `MiniMax-M2.7` | Minimax | ✓ | ✓ | minimax-m2.7, minimax-m2-7, minimax-m-2.7, minimax-m-2-7 |
439 | `MiniMax-M2.7-highspeed` | Minimax | ✓ | ✓ | minimax-m2.7-highspeed (+ dash variants) |
440 | `MiniMax-M2.5` | Minimax | ✓ | ✓ | minimax-m2.5 (+ dash variants) |
441 | `MiniMax-M2.5-highspeed` | Minimax | ✓ | ✓ | minimax-m2.5-highspeed (+ dash variants) |
442 | `MiniMax-M2.1` | Minimax | ✓ | ✓ | minimax-m2.1 (+ dash variants) |
443 | `MiniMax-M2.1-highspeed` | Minimax | ✓ | ✓ | minimax-m2.1-highspeed (+ dash variants) |
444 | `MiniMax-M2` | Minimax | ✓ | ✓ | minimax-m2, minimax-m-2 |
445 | `deepseek-ai/DeepSeek-V4-Pro` | Deepinfra | ✓ | ✓ | deepseek-v4-pro, di-deepseek-v4-pro |
446 | `deepseek-ai/DeepSeek-V4-Flash` | Deepinfra | ✓ | ✓ | deepseek-v4-flash, di-deepseek-v4-flash |
447 | `fugu` | Sakana | ✓ | ✗ | sakana-fugu, sakana/fugu |
448 | `fugu-ultra-20260615` | Sakana | ✓ | ✓ | fugu-ultra, sakana-fugu-ultra |
449 | `LongCat-2.0` | LongCat | ✓ | ✓ | longcat, longcat-2.0 |
450 | `muse-spark-1.1` | Meta | ✓ | ✓ | muse-spark, muse |
451 | `grok-4.5` | Xai | ✓ | ✓ | grok, xai-grok-4.5 |
452 | `grok-4.3` | Xai | ✓ | ✓ | xai-grok-4.3 |
453 | `grok-build` | Xai | ✓ | ✓ | xai-grok-build |
454 | `grok-composer-2.5-fast` | Xai | ✓ | ✗ | xai-grok-composer |
455 | `grok-4.20-0309-reasoning` | Xai | ✓ | ✓ | xai-grok-reasoning |
456 | `grok-4.20-0309-non-reasoning` | Xai | ✓ | ✗ | xai-grok-fast |
457
458 Resolution order (`resolve()` at `lib.rs:987+`): Ollama passes names through
459 verbatim; a `provider_hint` narrows the search; Atlascloud/Arcee/XiaomiMimo
460 accept arbitrary ids for their provider; otherwise falls back to the hinted
461 provider's first model, ultimate default `deepseek-v4-pro`.
462
463 ## 8. Fast-tier derivation contract
464
465 Auto routing and the legacy `model_strength = faster` (now the Scout role) use the provider-scoped family map
466 in `crates/tui/src/model_routing.rs`. A pair is valid only when the current
467 provider's catalog has both model ids; model-name similarity alone is not
468 enough. Current non-DeepSeek pairs are:
469
470 | Provider | Strong family | Fast sibling |
471 |---|---|---|
472 | OpenAI / OpenAI Codex | `gpt-5.6`, `gpt-5.6-sol`, `gpt-5.6-terra` | `gpt-5.6-luna` |
473 | Anthropic | Claude Opus/Sonnet rows listed in §4.3 | `claude-haiku-4-5` |
474 | OpenRouter | Qwen 3.6, Claude, MiMo, Trinity, Kimi rows | provider-prefixed catalog sibling |
475 | Xiaomi MiMo | `mimo-v2.5-pro` | `mimo-v2.5` |
476 | Arcee | `trinity-large-thinking` / `trinity-large-preview` | `trinity-mini` |
477 | Moonshot | `kimi-k2.7-code` | `kimi-k2.6` |
478 | MiniMax | `MiniMax-M2.7` | `MiniMax-M2.7-highspeed` |
479 | OpenCode Go | `kimi-k3` | `kimi-k2.7-code` |
480
481 Already-fast rows and true singles return no further sibling. Local/custom
482 providers never inherit cloud families. New catalog multi-tier families must
483 add an explicit same-provider pair and a matrix test before they become an
484 Auto fast route.
485
486 ## Appendix B — file map (where each fact lives)
487
488 | Fact | Primary file |
489 |---|---|
490 | Context window / max output / reasoning (static) | `crates/tui/src/models.rs:239-554` |
491 | Seeded metadata registry (future chokepoint) | `crates/tui/src/model_registry.rs` |
492 | Pricing (USD + DeepSeek CNY, time-aware rows) | `crates/tui/src/pricing.rs:112-303` |
493 | Bundled TUI catalog (31 entries, modalities) | `crates/tui/assets/model_catalog.bundled.json` |
494 | Bundled Models.dev seed (14 providers / 42 offerings) | `crates/config/assets/models_dev.bundled.json` |
495 | Live Models.dev cache (disk) | `~/.codewhale/catalog/models-dev-catalog.json` |
496 | Codex OAuth roster (disk, read-only) | `~/.codex/models_cache.json` |
497 | Provider descriptors (33) | `crates/config/src/provider.rs` |
498 | Provider base-URL/model constants | `crates/config/src/provider_defaults.rs` |
499 | Provider enum + parsing | `crates/config/src/provider_kind.rs`, `crates/tui/src/config.rs:42` |
500 | Catalog merge/precedence | `crates/config/src/catalog.rs`, `crates/tui/src/provider_lake.rs` |
501 | Live fetch/TTL/freshness | `crates/tui/src/models_dev_live.rs` |
502 | OAuth route limit override | `crates/tui/src/route_runtime.rs:33-51`, `crates/tui/src/codex_model_cache.rs` |
503 | Picker display precedence | `crates/tui/src/tui/model_picker.rs:1105-1155` |
504 | Agent-crate model×provider registry | `crates/agent/src/lib.rs:69-960` |
505
505 lines MARKDOWN