gpt-oss-120b

Description

"gpt-oss-120b" is a 117-billion-parameter open reasoning model designed for agentic intelligence, adaptive decision-making, and multi-step problem solving. It integrates structured reasoning with dynamic tool orchestration, enabling autonomous agents to plan, evaluate, and act across complex real-world contexts. Fine-tuned on diverse analytical and interaction-rich datasets, it maintains coherence over long reasoning chains, balances exploration with precision, and delivers transparent, goal-directed thought processes suited for research, operational, and strategic domains.

This model has the lifecycle status LTS. If we shut it down or replace it, we announce that at least 6 months in advance.

The following limitations apply:

• Maximum context length: 131,072 tokens
• "built-in tools" like web browsing not supported for security reasons. Standard tool-calling remains available.

Recommended inference parameters:

• temperature=1.0
• top_p=1

You can adjust the reasoning effort done by this model by defining it inside your system prompt in the following way: "Reasoning: high". You have three options:

• high: Deep and detailed analysis
• medium: Balanced speed and detail
• low: Fast responses for general dialogue

Terms of use and licensing

The model is provided by OpenAI under the Apache 2.0 License, and reuse of the generated content is not subject to any additional restrictions.