Command A+
Cohere's most powerful LLM, optimized to run on as little hardware as possible.
Cohere's open-source multimodal mixture-of-experts model release for agentic workloads, described as having 218B total parameters and 25B active parameters and deployable on two H100 GPUs.
Pricing
Input rate is explicitly for cache-miss input tokens; no separate cache-hit input rate was stated in the cited source.
The August 2026 third-party pricing page reports Cohere's hosted Command A+ API at $2.50 per 1M cache-miss input tokens and $10.00 per 1M output tokens. Cohere's official model documentation confirms text and image inputs with text output, and its pricing documentation confirms Command A+ is token-priced; the official pricing page did not expose a current numeric Command A+ row in the readable content.