Feature Parity Matrix
Use this page to see what you can migrate to AI Stats confidently today, and which areas still need extra care.
This summary is based on the code, tests, SDKs, and docs currently shipped in this repo.
| Capability | OpenRouter | Vercel AI Gateway | LLMGateway | Current AI Stats position |
|---|
| Public migration guide | Yes | Yes | Yes | Public docs cover all three migration paths |
| Base URL + API key swap | Yes | Yes | Yes | Covered in migration guides and SDK examples |
| Model-ID validation guidance | Yes | Yes | Yes | Covered via /v1/models verification guidance |
| Batch and file migration path | Partial | Partial | Partial | API, docs, and maintained SDKs cover batch create/retrieve/cancel, owned files, and result download |
| Async jobs and webhook visibility | Partial | Partial | Partial | Usage UI, websocket docs, and SDK helpers now cover batch/video async status and websocket URLs |
| Routing and pricing metadata | Strong | Medium | Medium | Usage UI, devtools, async jobs, and request views expose request IDs, provider attempts, and pricing lines |
| Vercel AI SDK compatibility proof | N/A | Strong | N/A | Local compatibility tests cover language, streaming, embeddings, image, speech, and transcription flows |
| Side-by-side benchmark tooling | Strong | Medium | Medium | Internal compare tooling now covers AI Stats vs OpenRouter, LLMGateway, and Vercel AI Gateway, with aggregate and live compare support when configured |
Product capability matrix
| Capability | Current public evidence | Status |
|---|
| Routing transparency | Usage UI, request views, async job details, devtools, and gateway error docs | Strong |
| Pricing transparency | Request pricing lines, batch settlement lines, async-job pricing metadata, pricing calculation endpoint, SDK helpers | Strong |
| Async and batch ergonomics | Batch create/retrieve/cancel, async websocket helpers, webhook-aware usage views, batch/video job docs | Strong |
| Provider/model filtering depth | /models now supports provider, status, rollout, availability, routing, and capability filters | Strong |
| Fallback controls for migrations | Public docs now point to preset constraints, routing mode, preview-channel toggles, and BYOK fallback controls as the supported migration knobs | Strong |
| Replay and retry ergonomics | GET /v1/generations?id=<request_id> now returns replay_supported plus a replay_request payload that can be resent to the original endpoint | Strong |
| Request transforms and defaults | Presets guide now defines the supported public controls for prompt injection, parameter defaults, reasoning defaults, and provider constraints | Strong |
| Broader competitor benchmarking | Operational compare flow now covers OpenRouter, LLMGateway, and Vercel AI Gateway; the remaining work is keeping that evidence aligned as contracts evolve | Strong |
What is already strong
- Migration guides now cover OpenRouter, Vercel AI Gateway, and LLMGateway.
- Maintained SDKs and docs cover batch, file, async-job, and gateway discovery helpers much more evenly than before.
- The dashboard and request logs now preserve much better routing, provider-attempt, and pricing context.
- Vercel AI SDK compatibility is now proven across the shipped text, embedding, image, and audio endpoints.
What still needs attention
- Keep the migration docs and comparison notes current as the product changes.
Last modified on June 10, 2026
