Command A+
Cohere · May 2026
● activeOpen Sourcemixture of expertsmultimodalAPI Available
Parameters218B (25B active)
Context Window256K tokens
VariantsCommand A+
Why It Matters
Largest fully open-source enterprise model at release, proving that frontier-class multimodal models could be both powerful and completely open.
Description
Cohere's most capable model, combining text and vision understanding in a sparse mixture-of-experts design (218B total parameters, 25B active). Supports 48 languages and is fully open-source under Apache 2.0, enabling deployment in government, on-premises, and air-gapped environments.
Notable Milestones
- ▸Sovereign AI deployments for governments
- ▸On-premises enterprise AI in regulated industries
- ▸Vision-enabled agentic workflows
Key Innovations
Reasoning
ReasoningStructured step-by-step problem solving, often using chain-of-thought or tree-of-thought approaches.
Multimodal
MultimodalProcessing multiple types of input (text, images, audio, video) in a single model.
Agentic
AgenticModels that can autonomously plan, execute multi-step tasks, use tools, and self-correct without human intervention.
MoE
MoEArchitecture where only a fraction of the model's parameters are active for each input, allowing massive scale with lower compute.