Key Points
- OpenAI has released two new “open-weight” AI models. The models are small enough to run on laptops and single GPUs.
- They are designed to excel in advanced reasoning, particularly in math and coding.
- This is OpenAI’s first open model release since 2019, re-entering a competitive space.
- Their performance is said to be similar to OpenAI’s smaller, closed models.
OpenAI, the company behind ChatGPT, has just released two new powerful AI models that are small enough for developers to run on their laptops. This is a major shift for a company best known for its massive, cloud-based AI systems.
The new models are “open-weight,” which means their trained parameters—the core of the AI’s knowledge—are publicly available. This allows developers to fine-tune the models for their own specific needs and run them locally, behind their firewalls, giving them more control and privacy.
“One of the things that is unique about open models is that people can run them locally,” OpenAI co-founder Greg Brockman said in a press briefing.
This is OpenAI’s first open model release since GPT-2 back in 2019, marking its return to a competitive field dominated by players like Meta’s Llama and China’s DeepSeek.
The company released two versions. The larger one, gpt-oss-120b, can run on a single powerful graphics card (GPU). The smaller model, gpt-oss-20b, is efficient enough to run directly on a personal computer.
OpenAI says the models are excellent at advanced reasoning, especially for tasks involving coding, math, and health-related questions. Their performance is reportedly similar to OpenAI’s own smaller proprietary models. The company trained them on a massive dataset focused on science, math, and coding.
