Command A+
Cohere · Released 20 May 2026 · command-a-plus-05-2026
–
Input price / 1M tokens
–
Output price / 1M tokens
128K
Context window
#70 of 71
211
Output tokens / second
#7 of 61
85%
τ²-bench
#4 of 6
63%
MMMU-Pro
#9 of 12
52.7%
CharXiv Reasoning
#6 of 6
Summary
Command A+ is an open-weight model from Cohere, released on 20 May 2026.
Its 128K-token context window ranks #70 of 71.
Measured output speed is 211 tokens per second, #7 of 61.
Cohere's first mixture-of-experts model (25B active, 218B total) and last of the Command A family, combining vision input, agentic tool use, reasoning and translation across 48 languages; it runs on 2 H100s or 1 B200.
Benchmarks · 6 reported
| Benchmark | Score | Bar | Rank | Source |
|---|---|---|---|---|
| τ²-benchstandard settings and user simulation | 85% | #4 of 6 | Cohere launch postSelf-reported | |
| Terminal-Bench HardTerminus-2 harness, Artificial Analysis methodology | 25% | – | Cohere launch postSelf-reported | |
| MMMU-Pro | 63% | #9 of 12 | Cohere launch postSelf-reported | |
| MMMU | 75.1% | – | Cohere launch postSelf-reported | |
| MathVista | 80.6% | – | Cohere launch postSelf-reported | |
| CharXiv Reasoning | 52.7% | #6 of 6 | Cohere launch postSelf-reported |
Self-reported means the lab ran the test itself. A rank appears only where several labs report the same version of a benchmark.
Similar models
Models with the most benchmarks in common with Command A+, and the closest scores.
τ²-bench
#4 of 6Tool use with a simulated customer
- Command A+85
- Gemma 4 31B76.9
- Gemini 3.1 Pro90.8
- Gemini 3 Flash90.2
- Gemma 4 26B A4B68.2
MMMU-Pro
#9 of 12Hard questions that need an image
- Command A+63
- Gemma 4 31B76.9
- Gemini 3.1 Pro80.5
- Gemini 3 Flash81.2
- Gemma 4 26B A4B73.8
- Gemini 3.5 Flash83.6
Compare Command A+ side by side
Pre-filled with the two closest models. Swap any of them.
Sources
- API ID, status, modalities, 128k context, 64k max output: docs.cohere.com/docs/models, checked 2 Oct 2026
- release date, license, size, input/output modalities: docs.cohere.com/docs/command-a-plus, checked 2 Oct 2026
- launch post: snapshot table and benchmarks: cohere.com/blog/command-a-plus, checked 2 Oct 2026
- pricing page lists Command A+ as Free; legacy Command prices: cohere.com/pricing, checked 2 Oct 2026
- trial and production rate limits: docs.cohere.com/docs/rate-limits, checked 2 Oct 2026
- trial usage is free: docs.cohere.com/docs/how-does-cohere-pricing-work, checked 2 Oct 2026
- output speed: artificialanalysis.ai/models/command-a-plus, checked 2 Oct 2026
Building on Command A+?
Get the architecture right before the bill arrives.
We size caching, routing and fallbacks for your workload on a free call, and tell you where a cheaper model is good enough.