GPT-OSS is a family of open source language models launched by OpenAI, which mainly contains two models: gpt-oss-120b and gpt-oss-20b. gpt-oss-120b has 117 billion parameters and is suitable for running in data centers or high-end devices, which can be driven by a single Nvidia H100 GPU; gpt-oss-20b has 21 billion parameters, optimized for low-latency scenarios, and can run on 16GB RAM devices. Both models are available under the Apache 2.0 open source license, allowing developers to freely download, modify and deploy them.
The main features include: support for chained reasoning (in low/medium/high strength), tool invocation (e.g., web search/code execution), structured output (Harmony format), 128k-long context processing capability, and model security through secure training and external auditing.
This answer comes from the articleGPT-OSS: OpenAI's Open Source Big Model for Efficient ReasoningThe
English 
简体中文 
日本語 
Deutsch 
Português do Brasil