Qwen3 235B, not served here yet
Qwen3-235B-A22B is the largest mixture-of-experts in Alibaba's Qwen3 family: 235B parameters, 22B active per token, Apache 2.0. OpenRelay does not serve it yet; tell us if you want it.
- Text in, text out
- Not served yet
- No public rate
Not served yet.
GET /v1/models lists Qwen3 235B with status request_access, but nothing serves it: there is no backend behind the id and no public rate. A request tells us someone wants it.
openrelay/qwen3-235bTell us you want Qwen3 235B
Leave a work email. We email you if we start serving it, and nothing else goes to this list.
The spec sheet.
What Alibaba has published about the model, with the source on each row.
- On OpenRelay
- openrelay/qwen3-235b: Text in, text out. Listed with status request_access, not served. GET /v1/models
- Parameters
- 235B total, 22B activated; 128 experts, 8 active Qwen model card
- Context
- 32,768 tokens native and 131,072 with YaRN in the original checkpoint Qwen model card
- Context, Instruct-2507
- 262,144 tokens native in the Instruct-2507 update, which is non-thinking only Qwen Instruct-2507 model card
- Modes
- The original checkpoint switches between thinking and non-thinking modes in one model Qwen model card
- Languages
- 100+ languages and dialects Qwen model card
- License
- Apache 2.0, open weights Qwen model card
Qwen3 235B, answered.
Can I call Qwen3 235B through OpenRelay?
Not today. openrelay/qwen3-235b appears on GET /v1/models with status request_access, but nothing serves it: a chat completions call returns 403 model_access_required. The form on this page tells us you want it.
Which Qwen3 235B release is this?
The catalog entry names the family, openrelay/qwen3-235b, not a release. Qwen published the original checkpoint, with thinking and non-thinking modes, and the Instruct-2507 update, non-thinking only with a longer native context.