spb/chat-spboucher-ai

1# Routing recommendations23Derived mechanically from `capability-matrix.json` (2026-11-01T07:22:05.282Z). The in-app router uses the same data at runtime; the owner can override any choice.45## Modes67- **Quality** — highest quality score; ties broken by higher list price (assumption: the vendor's higher tier is stronger on tasks harder than this suite), then latency; *-pro models excluded (latency): `claude-fable-5-1` (100%, 1321 ms, $10 / $50), `gpt-6-astra` (100%, 2190 ms, $10 / $50), `claude-fable-5` (100%, 2691 ms, $10 / $50)8- **Fast** — lowest median latency among models scoring ≥ 89 %: `gpt-4.1` (89%, 704 ms, $2 / $8), `claude-haiku-5-5` (100%, 883 ms, $0.1 / $0.5), `gpt-5.4-mini` (89%, 888 ms, $0.75 / $4.5)9- **Economy** — cheapest blended price among models scoring 100 %: `gpt-5-nano` (100%, 3034 ms, $0.05 / $0.4), `claude-haiku-5-5` (100%, 883 ms, $0.1 / $0.5), `gpt-6-luna` (100%, 2005 ms, $0.1 / $0.5)10- **Automatic** — classify the task (code / math / vision / long document / general / French), pick the first candidate of the matching task list below that is available, fall back to the Quality list.1112## By task1314- **code**: `gpt-5-nano`, `claude-haiku-5-5`, `gpt-6-luna`15- **math**: `gpt-5-nano`, `claude-haiku-5-5`, `gpt-6-luna`16- **reasoning**: `gpt-5-nano`, `claude-haiku-5-5`, `gpt-6-luna`17- **french**: `gpt-5-nano`, `claude-haiku-5-5`, `gpt-6-luna`18- **vision**: `gpt-5-nano`, `claude-haiku-5-5`, `gpt-6-luna`19- **tools**: `gpt-5-nano`, `claude-haiku-5-5`, `gpt-6-luna`20- **json**: `gpt-5-nano`, `claude-haiku-5-5`, `gpt-6-luna`21- **longctx**: `gpt-5-nano`, `claude-haiku-5-5`, `gpt-6-luna`2223## Specialized2425- Image generation: `chatgpt-image-latest`, `gpt-image-1-mini`, `gpt-image-1.5`, `gpt-image-2`, `gpt-image-2.5-flare`, `gpt-image-2.5-sunburst`26- Text-to-speech: `gpt-4o-mini-tts`, `tts-1`, `tts-1-hd`27- Speech-to-text: `gpt-4o-mini-transcribe`, `gpt-4o-transcribe`, `gpt-4o-transcribe-diarize`, `gpt-transcribe`, `whisper-1`28- Embeddings (memory & semantic search): `text-embedding-3-large`, `text-embedding-3-small`, `text-embedding-ada-002`2930## Defaults proposed for the app3132- Default chat model: `claude-fable-5-1`33- Fast mode: `gpt-4.1`34- Economy mode: `gpt-5-nano`35- Title generation / memory extraction / classification (cheap, fast): `gpt-5-nano`36- Embeddings: `text-embedding-3-small` (1536 dims, verified)37- Image generation: `gpt-image-1.5`38- STT: `gpt-4o-mini-transcribe` · TTS: `gpt-4o-mini-tts`39