gpt-oss

A pair of open-weightopen-weight model, n.: An AI model whose trained internal numbers are published for anyone to download and run on their own computers, even if the data and code used to make it stay private. reasoningreasoning model, n.: An AI model trained to work through a problem step by step in hidden 'thinking' text before it answers. Letting it think longer generally makes it better at math, coding and planning. language models released by OpenAIOpenAI, prop. n.: The American company that makes ChatGPT and the GPT models. It began in 2015 as a nonprofit research lab and is now a for-profit company overseen by a nonprofit foundation. on 5 August 2025 under the Apache 2.0 license: gpt-oss-120b, with about 117 billion parameters of which 5.1 billion are active for each token, and gpt-oss-20b, with about 21 billion parameters and 3.6 billion active. Both are mixture-of-expertsmixture of experts, n.: A way to build very large AI models cheaply: the model contains many specialist sub-networks, and for each word only a few are switched on, picked by a small dispatcher network. transformerstransformer, n.: The design behind today's chatbots and much of modern AI. It handles text by letting every word look at every other word and decide which of them matter for understanding it. whose expert weights are stored in the 4-bit MXFP4 format (see FP4FP4, abbr.: A tiny 4-bit number format that can represent only sixteen different values. With clever scaling tricks it lets the newest AI chips run, and even train, models faster and with less memory.), so the larger fits on a single 80 GB GPU and the smaller runs within 16 GB of memory. They offer a 128,000-token context, adjustable reasoning effort, tool usetool use, n.: An AI model's ability to reach beyond its own knowledge by operating outside tools — searching the web, running code, using a calculator or an app — and folding the results into its work. such as web browsing and code execution, and a chat template called the harmony format. OpenAI reported that the larger model came close to its o4-mini on reasoning benchmarks. Before release, OpenAI fine-tuned a version adversarially to estimate worst-case misuse and reported that it still did not reach the High capability threshold for biology or cybersecurity under its Preparedness FrameworkPreparedness Framework, prop. n.: OpenAI's published process for testing its AI models for catastrophic risks — such as help with biological weapons or cyberattacks — and deciding what safeguards are needed before release.. Its raw chain of thoughtchain of thought, n.: Getting an AI to write out its reasoning step by step before giving an answer, the way a student shows their working. This often makes answers to math and logic problems more accurate. was kept free of direct training pressure so that it can be monitored.

We run gpt-oss-20b on a laptop for offline coding help and send only the hard tasks to the 120b model.