Command A+
Cohere's fastest and most powerful open-weight model, aimed at high-performance enterprise agents: a 218B sparse mixture-of-experts architecture with native citations, tool use and reasoning, multilingual across 48 languages, released under an Apache 2.0 licence. It has three distinct cost paths, the weights are a free download and run on as little as two H100 GPUs, Cohere's trial API key is free but rate-limited and not licensed for production, and the hosted production API bills $2.50/$10 per MTok. The rate we publish is the production API, since that is what a live workload actually pays.
Command A+ is a frontier AI model from Cohere. It costs $2.500 per million input tokens and $10.000 per million output tokens (blended $4.750/M), with a 256,000-token context window.
- Apache 2.0 open weights, self-host with no per-token charge
- Runs on as few as two H100 GPUs
- Native citations for grounded answers
- Multilingual across 48 languages
- Tool use and reasoning for agents
- 218B sparse MoE for efficiency at scale
- Enterprise agents needing citations
- Self-hosted or sovereign deployments
- Multilingual retrieval and support
- Regulated environments where weights must stay in-house
Benchmarks
More from Cohere
See all 12 →Frequently asked questions
How much does Command A+ cost?
Command A+ costs $2.500 per million input tokens and $10.000 per million output tokens, for a blended reference rate of $4.750 per million tokens.
What is Command A+'s context window?
Command A+ supports up to 256,000 tokens of context in a single request.
What is Command A+ best for?
Command A+ is well suited to Apache 2.0 open weights, self-host with no per-token charge, Runs on as few as two H100 GPUs and Native citations for grounded answers.
Who makes Command A+?
Command A+ is developed and served by Cohere.