General Reasoning

Need a reasoning model you can actually inspect and build on? General Reasoning works on exactly that problem.

General Reasoning is an open source AI models project that develops reasoning AI through reinforcement learning, with the stated goal of making machines that are both smart and safe.

The team's focus is the frontier of machine reasoning.

A large part of that is scaling the compute spent on reinforcement learning, since that is where they see reasoning ability improving.

They also say they open-source their data and models, which matters if you want to study, fine-tune or reproduce the results instead of only calling a closed API.

General Reasoning describes its mission as making state-of-the-art reasoning AI more accessible, and the work is pitched as something that should help humanity broadly.

That is a large claim, and the site is the place to check what has actually been released.

Researchers and developers who care about open reasoning models and reinforcement learning methods are the natural audience.

If you want a polished consumer chat product, this is probably not that.