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

Official site · Aug 22, 2026, 1:10 PM

Input / 1M

$2.50

Output / 1M

$10.00

Production API pay-as-you-go pricing. Trial API-key usage is free but limited; Command A+ is also available as Apache-2.0 open weights for self-deployment.

Cohere's pricing documentation identifies Command A+ as a generative model billed per input and output token; Cohere's public Command A rate card lists USD 2.50 per 1M input tokens and USD 10.00 per 1M output tokens. Command A+ accepts text and image inputs and produces text.

View source

Model Intelligence

Arena ranking

Benchmarkable

Yes

Model level

release

Intelligence Index

22.8

Coding Index

27.8

GPQA

0.76

HLE

0.12

SciCode

0.38

IFBench

0.74

LCR

0.49

TerminalBench Hard

0.25

TAU2

0.81

Recent stories

1 linked story

releasePRIMARY2026-05-20

Cohere releases Command A+ under Apache 2.0 with 25B active params and 2x H100 deployment

Cohere open-sourced Command A+, a 218B MoE multimodal model with 25B active parameters, 48-language support, and deployment starting at two H100s. Artificial Analysis put it at 37 on its Intelligence Index and 281 tok/s, and vLLM plus Transformers added support.