Cohere · 2026-05-20 · major
Cohere Command A+ — 218B Sparse MoE With 25B Active Params, Apache 2.0, Runs Agentic and Multimodal Workloads on Two H100s
Cohere released Command A+, a 218B mixture-of-experts model with 25B active params under an Apache 2.0 license. It handles vision and 48 languages, runs on two H100s, and scores 85% on the tau2-Bench Telecom agentic benchmark.
An open-weight 218B MoE that runs agentic, multimodal, multilingual workloads on as few as two H100 GPUs.
Key specs
| Parameters | 218B |
|---|---|
| Active params | 25B |
| Context window | 128K |
| Languages | 48 |
| Tau2 bench telecom | 85% |
| Terminal bench hard | 25% |
What is it?
Command A+ is Cohere's new open-weight large language model, released under an Apache 2.0 license. It folds four earlier Command models — base, reasoning, vision, and translate — into one mixture-of-experts model aimed at enterprise agents that organizations can self-host for data sovereignty.
How does it work?
It uses a sparse mixture-of-experts design: 218B total parameters but only 25B activate per token, so a 4-bit W4A4 build fits on one B200 or two H100 GPUs. It supports a 128K-token input window and generates native citation grounding spans that tie each factual claim back to the source document or tool output.
Why does it matter?
Enterprises and public-sector teams that need data residency can run a capable agentic, vision-capable model on-prem without per-token API fees. Apache 2.0 lets them modify and redeploy the weights freely, and the two-H100 footprint keeps the hardware bill in reach.
Who is it for?
enterprise and public-sector AI teams
Try it
huggingface.co/CohereLabs/command-a-plus-05-2026-w4a4