Limits and compatibility
Check the endpoint and model before sending a large request. OpenAI compatibility does not imply support for every OpenAI method.
01
Supported
GET /v1/modelsPOST /v1/chat/completionsPOST /v1/responsesPOST /v1/messagesPOST /v1/images/generations/v1/cursor/*forgyw-opus-5
02
Not supported
Legacy completions, embeddings, rerank, moderations, audio, realtime, videos, image edits, and Gemini-native endpoints.
03
Baseline limits
Values can be overridden for an individual key.
| Limit | Baseline |
|---|---|
| Requests per key | 120 / minute |
| Concurrent requests per key | 8 |
| Global concurrency | 48 |
| Normal request body | 25 MiB |
| Images per request | 1 |
04
Model restrictions
The catalog endpoints field is the source of truth. For example, image models are not accepted by Chat Completions, and the Cursor bridge accepts only gyw-opus-5.