gpt-oss-120b
Open Weights
OpenAI
OpenAI's first open-weight language model since GPT-2 (August 2025), released under Apache 2.0. A 117B-parameter mixture-of-experts model with 5.1B active parameters per token, it reaches near-parity with o4-mini on core reasoning benchmarks while running on a single 80GB GPU thanks to MXFP4 quantization.
Strengths
- OpenAI's first open-weight model since GPT-2, under permissive Apache 2.0
- Near-parity with o4-mini on core reasoning benchmarks
- 117B-parameter MoE with only 5.1B active parameters per token
- Runs on a single 80GB GPU via MXFP4 quantization
- 128K context with strong tool use and agentic workflows
Caveats
- Text-only (no vision or audio input)
- Requires an ~80GB GPU for full-precision local serving
- Trails proprietary frontier models on the hardest tasks
- Vendor
- OpenAI
- Context Window
- 131,072 tokens
- License
- Apache-2.0
- Release Date
- 2025-08-05
- Modalities
- text
Capabilities
Vision
Audio
Video
Tool Use
Pricing
Open weights under Apache 2.0 - free to download and self-host; runs on a single 80GB GPU (H100/MI300X). Hosted API providers vary.