Revolutionary Open-Source Reasoning Models

Independent Open-Source Project

This project is not affiliated with, endorsed by, or connected to OpenAI. GPT-OSS models are independently developed open-weight language models created by the open-source community under Apache 2.0 license.

Discover GPT-OSS, the breakthrough open-source AI models with configurable reasoning levels, native agentic capabilities, and permissive Apache 2.0 licensing. Run on consumer hardware or scale to enterprise deployments.

120B
Parameters (5.1B active)
Apache 2.0
Open Source License
16GB
Minimum Memory (20B model)

Choose Your Model

Two powerful variants designed for different use cases and hardware requirements

GPT-OSS-120B

117B parameters (5.1B active)

Designed for production and general-purpose high reasoning use cases

Key Features

  • Configurable reasoning levels
  • Full chain-of-thought reasoning access
  • Native MXFP4 quantization
  • Runs on a single H100 GPU

Capabilities

  • Function calling
  • Web browsing
  • Python code execution
  • Structured outputs

GPT-OSS-20B

21B parameters (3.6B active)

Can run within 16GB of memory for consumer hardware

Key Features

  • Configurable reasoning levels
  • Full chain-of-thought reasoning
  • Native MXFP4 quantization
  • Fine-tunable on consumer hardware

Capabilities

  • Function calling
  • Web browsing
  • Python code execution
  • Structured outputs

Why Choose GPT-OSS?

Built from the ground up with cutting-edge features for modern AI applications

⚙️

Configurable Reasoning

Choose from low, medium, or high reasoning levels to match your specific use case

🤖

Agentic Capabilities

Built-in support for function calling, web browsing, and Python code execution

🛡️

Safety-First Design

Comprehensive safety measures and responsible AI practices built into every model

📜

Apache 2.0 License

Completely free for both commercial and non-commercial use with permissive licensing

🔗

Broad Ecosystem Support

Compatible with Transformers, vLLM, PyTorch/Triton, Ollama, and LM Studio

Optimized Performance

Native MXFP4 quantization for efficient inference on consumer hardware

5 Frameworks
Supported inference engines
MXFP4
Native quantization
Chain-of-Thought
Full reasoning access
Fine-tunable
Customize for your needs

Frequently Asked Questions

Everything you need to know about GPT-OSS models