product · inferred from evidence
qwen-3-235b-a22b-thinking-2507
A model that only emits a closing [/think] tag with no opening tag.
- Cerebras reasoning models put reasoning in content, not a reasoning key
With streaming on Cerebras, reasoning tokens arrive in the content field wrapped in think tags for most models. Only gpt-oss-120b uses a separate reasoning key. Parse the think blocks client-side.