Microsoft makes the open-source GPT model available for Windows 11 computers
Microsoft recently announced the integration of the gpt-oss-20b model, developed by OpenAI, into the Windows AI Foundry platform. Thanks to this change, Windows 11 users will be able to access this open-source AI model for free from their devices.
According to Microsoft, this model is ideal for independent tasks and resource-constrained environments.
The gpt-oss-20b model is described as "lightweight and easy to use." The company added that it is "optimized for algorithmic tasks, such as executing code and using tools. It runs efficiently on a variety of Windows computers, with support for more devices coming soon."
The model, which is already available, can run on PCs and laptops with at least 16 GB of VRAM, which is very common for modern graphics cards, such as those from NVIDIA or Radeon.
The model was originally trained using advanced, computationally intensive boosting techniques, allowing it to excel at controlling agents and tools such as web browsers or code execution.
However, the gpt-oss-20b model is text-only, so it cannot process or generate images or audio.
It also appears to have hallucination problems, as, according to OpenAI, in internal tests the model gave incorrect answers to 53% of questions on human knowledge.
However, Microsoft plans to expand its use to macOS and other devices in the future.
With this, both Microsoft and OpenAI are betting on models that are increasingly available on a wider range of platforms.