Model Provider Support#

Use when answering team questions about Claude, GPT/OpenAI, or Gemini API/model support across cloud-hosted providers.

Rules#

• Treat each provider-hosted model as its own product surface.
• Do not infer cloud provider support from first-party model docs, or first-party support from cloud provider docs.
• Cite official provider documentation for the exact surface being discussed.
• If docs conflict or are stale, say so and prefer the hosting provider's current docs for hosted availability.
• Mark unverified capabilities as unknown_needs_live_check; do not guess.

Workflow#

1. Identify model family, hosting provider, region, API surface, and capability.
2. Load references/source-registry.json and the relevant provider notes.
3. Check references/capability-taxonomy.md for status labels.
4. Apply references/provider-differences.md before answering.
5. If live docs cannot be reached, cite registry URL and access notes, then answer conservatively.

Validation#

Run node scripts/check-sources.mjs after editing registry entries.