gpt-oss-120b
- Name in the API
openai/gpt-oss-120b- Base model
- openai/gpt-oss-120b
- Weights we run
- The original weights (MXFP4, as published)
- Type
- Chat
- Input → output
- text → text (images in the request are ignored)
- Context length
- 131,072 tokens
- Availability
- Dedicated
- Hardware
- One GPU with at least 90 GB (NVIDIA H100 94 GB or RTX PRO 6000 96 GB)
- Thinking
- Default: always on, with effort
medium.
reasoning_effort: low, medium, high. none, minimal, xhigh and max are rejected with error 400.- The levels work as expected: in our tests
high thought three to four times as much as medium,
and low somewhat less than medium.
- Numeric effort: no.
- Turn off: not possible;
chat_template_kwargs switches have no effect. Use low for the shortest thinking.
- Tool calling
- Yes
- Example
- Chat request. Set
model to the name above.