Hi everyone!
Cohere has just open-sourced their new reasoning model, Command A Reasoning.
It’s a powerful model designed for private enterprise deployment. Users can control the token budget to switch between instant responses and deep reasoning with a single deployment, a very practical approach for managing GPU costs.
One thing I really like is that its thinking process isn’t hidden or abstracted away. The chain-of-thought is very detailed, which is great for enterprise apps that need auditable reasoning.
You can try it out on the Cohere Playground and HF Spaces.