Strengths
- Open weights under the permissive Apache 2.0 license, which allows commercial use and fine-tuning for your own tasks.
- Runs on a single 80 GB GPU such as an H100. OpenAI says it nearly matches o4-mini on core reasoning tests.
- You can set thinking to low, medium or high and see the full chain of reasoning. It also supports structured outputs and function calling.
Weaknesses
- Knowledge stops in June 2024 and it only understands text, so it cannot handle images or scanned documents.
- Its context window is 131K tokens, far less than the roughly one million of the GPT-6 API models.
- Trained mostly on English text, so it may be weaker in Slovak and Czech.
- Running it yourself takes an expensive GPU and someone to set up and maintain the model.
Best for
- on-premise use, even without an internet connection
- data that must not leave your own infrastructure
- fine-tuning for specialised tasks
Not ideal for
- current events and news
- images and scanned documents
- top-level reasoning on par with GPT-6
How to use it in Europe
It is unclear whether the OpenAI API still serves it: it is missing from the price list and the model page shows zero rate limits. So expect not to use it through OpenAI directly. You can download the weights for free from Hugging Face and run them on your own server, or use a provider that hosts the model. The Apache 2.0 license has no geographic limits, so you can run it on a server in the EU and keep your data in house. With a third-party provider, check where they process your data.
More models from this vendor
- GPT-6 Astra: OpenAI's most capable model, built for the hardest coding, agents that operate a computer and very long documents.
- GPT-6.1 Sol: OpenAI's balanced model, close to GPT-6 Astra in quality at a fifth of the price. A solid pick for business apps.
- GPT-6.1 Terra: OpenAI's fast, cheaper model for working with images, operating a computer and everyday office work.
- GPT-6 Luna: OpenAI's cheapest model, made for high volumes of simple tasks, sorting text and low-cost chatbots.
- gpt-oss-20b: OpenAI's smaller open model that runs on a computer with 16 GB of memory, even without internet access.
Similar models from other vendors
- Claude Fable 5.1 (Anthropic): Claude's most expensive model, meant for the hardest tasks and long autonomous work when even Opus 5.5 falls short.