11ai Vercel AI Gateway models and capabilities

A model string chooses capability, provider options, price, and data terms Resolve the exact project, environment, remote target, permission and cost boundary, and acceptance check before acting.

Version baseline: Use current generally available AI Gateway behavior as of July 2026, including AI SDK 7, multimodal and realtime models, routing, budgets, BYOK, regional inference, and data-retention controls.

Inspect first

rg -n 'model:|creator/model|ai-gateway.vercel.sh|/v1/models' TARGET
npx vercel ai-gateway --help 2>/dev/null | head -100

Resolve exact current model metadata from the public catalog or models endpoint and compare application requirements, including text, image, video, embeddings, speech, transcription, or realtime voice capability.

Confirm before changing:

• Canonical creator/model identifier.
• Required modality and tool or schema support.
• Context and output bounds.
• Current pricing and provider availability.

Operate

npm test --if-present -- TARGET
npm run typecheck --if-present

Centralize approved model aliases, validate capabilities before requests, and record catalog retrieval time without hardcoding transient availability.

Never switch production models or make benchmark calls without approval, budget, and behavior review Require explicit approval and preview targets, permissions, counts, cost, remote effects, and rollback.

Verify and report

npm run typecheck --if-present
npm test --if-present
npm run build --if-present

Test one bounded request or mock for each required capability and confirm usage metadata and unsupported-feature handling. Report scope, target IDs, files, remote actions, permissions, cost, checks, observability, and rollback. Hand configuration failures to 11ai-operator-vercel-ai-gateway-troubleshooting and seams to 11ai-operator-vercel-ai-gateway-integrations.