Cohere · AI model
Command A+
Cohere's first mixture-of-experts model (218B total / 25B active parameters), combining text and image input, tool use and switchable reasoning across 48 languages; 128K context, 64K output; weights under Apache 2.0.
Last verified Available 3 sources
Key specs
- Provider
- Cohere
- Family
- Command A
- API identifier
- command-a-plus-05-2026
- Availability
- Available
- Release date
- 2026-05-20
- License
- Open source (Apache 2.0)
- Context window
- 128,000 tokens
- Max output
- 64,000 tokens
- Knowledge cutoff
- April 2025
- API access
- Yes
- Consumer app
- No
Capabilities
Inputs
- Text input Yes
- Image input Yes Input modalities listed as text and image
- Audio input No Input modalities listed as text and image only
- Video input No Input modalities listed as text and image only
- File input Unknown
Outputs
- Text output Yes
- Image output No Output modality listed as text only
- Audio output No Output modality listed as text only
Features
- Reasoning Optional Reasoning listed in the capabilities panel on the model's docs page; the page refers to the Reasoning guide for enabling/disabling thinking and token budgets (Cohere reasoning models are hybrid, thinking on by default)
- Tool calling Yes Tool Use listed in the capabilities panel on the model's docs page; tool use guide examples use command-a-plus-05-2026
- Web search Unknown
"Unknown" means we have not verified this capability yet; it does not mean the model lacks it.
Pricing
No current prices listed.
Benchmarks
| Benchmark | Setting | Score | Reported by | Measured | Verified | Source |
|---|---|---|---|---|---|---|
| CharXiv Reasoning accuracy %, higher is better | CharXiv reasoning questions; tools, split and thinking mode not stated ('standard methodology') | 52.7 % | Self-reported | — | 2026-10-25 | Introducing Command A+ | Cohere |
| GDPval-AA v2.1 Elo, higher is better | GDPval-AA v2.1; effort not stated (row label 'Command A+'); Stirrup agent harness with shell and web browsing | 514 Elo | Third party | 2026-10-25 | 2026-10-25 | GDPval-AA v2.1 Leaderboard (Artificial Analysis) |
There is no maintained public leaderboard (the project site says it is under construction), so almost all current numbers are self-reported by providers. Results with tools such as a code interpreter are much higher than without: Alibaba reports 83.7% without and 90.2% with tools for Qwen3.8-27B. Top models now exceed the human baseline, so differences at the top are small and may reflect label noise. Descriptive-question scores are a different metric and must not be recorded here.
Elo is a relative rating: values change as models are added and the scale is re-anchored, so always store the date a rating was read. For example, Anthropic's launch page gives 1846 for Claude Opus 5.5, while the leaderboard on 2026-10-12 shows 1866 (Max, Default Fallback). Earlier versions (Google cites "GDPVal-AA v2") are not comparable. Judging is done by an LLM, not by human experts. Anthropic models are listed with "Default Fallback", meaning refused steps are routed to a fallback model. AA-Briefcase and GDP.pdf are separate evaluations.
One row per result, without rank numbers: each result was measured under its own setting (effort level, harness, tools, test version), as stated by the source, and scores compare only within the same setting. Self-reported results come from the provider and have not been reproduced independently.
Compare Command A+
Guides
About Command A+
Command A+ is a sparse mixture-of-experts model with 218B total and 25B active parameters, released on May 20, 2026; Cohere's release notes describe it as the last model in the Command A family. It combines image input, reasoning, translation and agentic tool use in one model and supports 48 languages, including all official EU languages. Weights are published under Apache 2.0, with 1x B200 or 2x H100 at W4A4 listed as required hardware. On the Cohere API it is free for trial and production keys up to the rate limits, and production use runs through Model Vault.
Strengths and weaknesses
Strengths
- Accepts text and image input, with a 128,000-token context window and up to 64,000 output tokens
- Supports tool use, structured outputs and hybrid reasoning (thinking can be enabled, disabled or limited with a token budget)
- Supports 48 languages, including all official EU languages
- Open weights under Apache 2.0; 1x B200 or 2x H100 (W4A4) listed as required hardware
Weaknesses
- No per-token API price is published; Cohere API use is free only up to the rate limits (20 requests/min, 1,000 calls/month), and production use requires Model Vault or contacting sales
- Smaller context window (128K) than Command A and Command A Reasoning (256K)
Sources
Last verified Page updated 2026-10-26