Text and chat · OpenAI
GPT OSS
OpenAI's open-weight model, run in private mode: cheap, quick, with adjustable reasoning. Good for everyday questions.
Private Low costWhat sets it apart
gpt-oss is OpenAI’s open-weights model: 117B parameters with only 5.1B active, small enough for a single 80 GB GPU, under the permissive Apache 2.0 license. It is older and far less capable than GPT-6, but very fast and cheap to run.
- Apache 2.0 open weights
- 117B MoE, 5.1B active
- Low, medium or high reasoning
Strengths and limits
Where it shines
- Very low price, private mode
- Adjustable reasoning depth
Think twice when
- Text only, 128K context
- The encrypted version has no tool calling, so none of Sub Rosa's agent tools
- 117B-parameter mixture of experts
- 128K token context
Use cases
Quick answers at low cost
Turn these meeting bullets into a short follow-up email to the client, polite and under 120 words.
It generates about 190 tokens per second, per Artificial Analysis.
Structured extraction
From this invoice text, extract supplier, date, total and VAT as a JSON object.
It supports function calling and structured output.
Against its rivals
- GemmaGemma 4 31B scores 15 on the index against 12, and also accepts images.#71 of 90 on Artificial Analysis Intelligence Index (17 points)
- Qwen open modelsQwen’s open line goes much higher (Qwen3.8 27B reaches 34) and offers more sizes.#23 of 90 on Artificial Analysis Intelligence Index (40 points)
- GLMGLM 5.3 Flash scores 42, far above, with a 1M context.#13 of 90 on Artificial Analysis Intelligence Index (45 points)
Version history
Each version, dated, with what it changed over the one before.
- gpt-oss-120bLatest
- First open-weights OpenAI model in the catalog: 117B mixture of experts, 131K context, adjustable reasoning effort, Apache 2.0.
Benchmarks
How to read these scores: what a benchmark tells you.
Artificial Analysis Intelligence Index: #79 of 90 models measured in the catalog. One score averaging about ten independent tests of reasoning, knowledge, math, coding and agent work, all run the same way for every model. Higher is better. Each model is listed at its highest reasoning effort. Two or three points apart is a close call; ten points is a different class of model.
See the numbers
| Model | points |
|---|---|
| 1. Claude Opus 5.5 | 58 |
| 2. Claude Sonnet 5.5 | 56 |
| 3. Claude Fable 5.1 | 53 |
| 4. GPT-6 Astra | 53 |
| 5. Gemini 4 Argon | 53 |
| 6. GPT-6.1 Sol | 52 |
| 7. Muse Wonder | 52 |
| 8. Claude Opus 5 | 51 |
| 9. Claude Fable 5 | 50 |
| 10. GPT-6 Sol | 48 |
| 79. OpenAI GPT OSS 120B | 12 |
Announced by the maker
Measured by the maker with its own settings: useful, but not directly comparable with the independent scores above.
| Benchmark | Model | Score | Source |
|---|---|---|---|
| gpqa-diamond | OpenAI GPT OSS 120B | 80.81 | huggingface.co |
All 2 models in this family
| Model | Use | Privacy | Price |
|---|---|---|---|
OpenAI GPT OSS 120Bopenai-gpt-oss-120b | Text | Private | $0.135 per million tokens written$0.0000944 a page written |
GPT OSS 120B E2EEe2ee-gpt-oss-120b-p | Text | Private | $0.292 per million tokens written$0.000205 a page written |
Specifications
| Specification | gpt-oss-120b |
|---|---|
| Released | August 2025 |
| Parameters | 117B total, 5.1B active |
| Weights | Open, Apache 2.0 |
| Accepts | text |
| Produces | text |
| Context | 128K tokens |
| Abilities | Uses Sub Rosa’s tools, Thinks before answering |
| Price | $0.135 per million tokens written |
| Privacy | Private |
Prices, context and lengths from the live catalog on October 16, 2026; parameters and licenses from the maker.
Sources
Everything on this page comes from these pages, read on October 16, 2026 or as dated on each score.