Cohere Launches Command A+ – Fast, Efficient MoE Model
Integrate Command A+ into your agent workflows for speed and low GPU usage.
Add Command A+ to your agent stack to improve performance.
Summary
Cohere announces Command A+, its first Mixture‑of‑Experts model, designed to be fast and responsive. The model runs efficiently on 1–2 GPUs thanks to aggressive quantization, making it practical for smaller teams. Command A+ is released under Apache 2.0, providing near‑unrestricted access for developers.
While top‑line performance is still improving, the model’s speed and low GPU footprint make it attractive for agent platforms. Cohere emphasizes its enterprise focus but encourages community feedback to shape future iterations. The release marks a significant step toward making large‑scale LLMs accessible to smaller organizations.
Key changes
- First MoE model from Cohere.
- Fast and responsive.
- Runs on 1–2 GPUs due to quantization.
- Released under Apache 2.0 license.
- Designed for small teams.
- Open‑source.
- Performance still improving.
- Supports agent platforms.