AI models

WebCanon’s AI resolver passes the model id straight through to the provider’s SDK, so any model the provider offers works — set it with WEBCANON_AI_MODEL or --ai-model. The tables below list representative, currently-available model strings per provider; bold is WebCanon’s per-provider default.

The authoritative, always-current list lives in each provider’s docs (linked per section). Model ids — especially OpenAI’s and Gemini’s — change often. WebCanon does not validate the string: an unknown/retired id makes the resolver decline and fall back to the rule engine (retrieval still succeeds). Check inspect → LLMS: resolved-by (ai vs rule_based) to confirm the AI ran.

Picking a model for this task

The AI resolver does one small job per fetch: read the URL + llms.txt + robots verdict and return the URL to fetch and any safe headers. It’s a short, low-stakes tool call — a fast/cheap model is usually plenty.

  • Cost / latency first (recommended): gpt-4o-mini, gemini-2.5-flash, claude-haiku-4-5
  • Maximum quality: the per-provider defaults below
webcanon fetch https://example.com/docs/api --ai-provider openai --ai-model gpt-4o-mini

Anthropic (WEBCANON_AI_PROVIDER=anthropic)

Key: ANTHROPIC_API_KEY · Extra: pip install "webcanon[ai]" · Docs: https://docs.anthropic.com/en/docs/about-claude/models

Model id Notes
claude-fable-5 Most capable; hardest reasoning / long-horizon agentic work
claude-opus-4-8 (default) Top Opus-tier; highly autonomous
claude-opus-4-7 Previous-generation Opus
claude-sonnet-4-6 Balanced speed and intelligence
claude-haiku-4-5 Fastest and most cost-effective

OpenAI (WEBCANON_AI_PROVIDER=openai)

Key: OPENAI_API_KEY · Extra: pip install "webcanon[openai]" · Docs: https://platform.openai.com/docs/models

Model id Notes
gpt-5 (default) Flagship general model
gpt-5-mini Smaller, lower-cost gpt-5
gpt-4o Multimodal general-purpose
gpt-4o-mini Small, low-cost gpt-4o (good default for this task)

Custom / OpenAI-compatible endpoints: set OPENAI_BASE_URL or pass OpenAiAiResolver(base_url=...) in code.

Google Gemini (WEBCANON_AI_PROVIDER=gemini)

Key: GEMINI_API_KEY (or GOOGLE_API_KEY) · Extra: pip install "webcanon[gemini]" · Docs: https://ai.google.dev/gemini-api/docs/models

Model id Notes
gemini-2.5-pro (default) Top tier; long context, hardest tasks
gemini-2.5-flash Fast, low-cost, balanced
gemini-2.5-flash-lite Lightest and cheapest

Where the defaults live

The per-provider defaults are defined in webcanon.ai.DEFAULT_MODELS and used whenever WEBCANON_AI_MODEL / --ai-model is unset. See customization.md for provider/key/flag details.