OpenAI releases open‑weight GPT models that run on laptop

OpenAI has unveiled two open‑weight language models designed to deliver advanced reasoning capabilities on consumer hardware, including laptops and desktops. Released August 5, these are OpenAI’s first models with publicly available parameters since GPT‑2 in 2019, marking a strategic pivot toward transparency and accessibility. 

The two models are known as gpt‑oss‑120b and gpt‑oss‑20b. The larger model, gpt‑oss‑120b, can run on a single GPU, while the smaller gpt‑oss‑20b can operate on machines with at least 16 GB of memory. Both demonstrate performance comparable to OpenAI’s proprietary o3‑mini and o4‑mini reasoning models, particularly in coding, math, and advanced logic tasks. 

Unlike open‑source models, open‑weight means that only the trained parameter weights are released—developers can download and fine‑tune the models, but training data and full code remain closed. OpenAI emphasized safety by conducting internal misuse simulations and engaging external experts before releasing the models. 

This launch makes OpenAI’s models more readily available on local devices, offering cost savings, privacy benefits, and greater flexibility for developers. OpenAI co‑founder Greg Brockman said that users can run the models behind a firewall on their own infrastructure, without relying on cloud-based services. 

Amazon’s AWS Bedrock has also announced it will support these open‑weight models, making them available through the Bedrock generative AI marketplace. This marks the first time OpenAI’s weights are being offered via AWS, signalling deeper collaboration across platforms. 

For the Indian market, where cost‑efficient and offline AI deployment is critical, GPT‑OSS offers new possibilities. Smaller developers, educational institutions, and enterprises with limited cloud budgets can now run sophisticated AI workflows locally. This could accelerate local innovation in content creation, coding tools, and domain‑specific AI applications.

The move also reflects intensifying competition in the global open‑weight AI space, amid earlier successes from China’s DeepSeek and Meta’s Llama models. OpenAI framed the release as a step toward democratizing AI tools, highlighting its alignment with democratic values and broader vision of AGI for humanity. 

On the product front, the models are limited to text-based tasks—they do not support images, audio or video like some of OpenAI’s recent multimodal offerings. However, they can access the web, operate as agents, execute code, and perform chain‑of‑thought reasoning. 

This announcement comes ahead of an anticipated GPT‑5 rollout, expected to integrate OpenAI’s o‑series models into a unified system offering broader functionality.