Editorial illustration for Cohere open-sources Command A+ with 128k token window, 48‑language support
Cohere open-sources Command A+ with 128k token window,...
Cohere has thrown open the doors to its most capable model yet. Command A+ arrives with a 128,000-token context window, support for 48 languages, and the ability to process both text and images, all packed into a single open-source release. The performance numbers tell a stark story: on the agent benchmark τ²-Bench Telecom, scores vaulted from 37 to 85 percent; on the coding test Terminal-Bench Hard, they jumped from a meager 3 to 25 percent.
That’s not incremental, it’s transformational. On the Artificial Analysis Intelligence Index, Command A+ lands just under 37 points, putting it in the same league as Claude 4.5 Haiku, Gemma 4 31B, and Mistral Medium 3.5. Weights are already live on Hugging Face in multiple quantizations.
For enterprises building agentic workflows, RAG pipelines, or multilingual document processing, this isn’t just another model drop, it’s a new baseline.
This isn’t just another open-weight release. Command A+ lands with a 128k context window, real multilingual fluency, and benchmark scores that leap, not creep, ahead of its predecessor. On agent tasks, on coding, on the kind of messy, multi-step workflows enterprises actually run.
The weights are on Hugging Face. The quantizations are ready. And against peers like Claude 4.5 Haiku or Gemma 4 31B, it holds its ground without apology.
Cohere didn’t hoard its strongest model. They handed it over. What happens next depends entirely on what builders do with it.
Related Reading
Open Source Trump cracks down on Anthropic after Amazon tip; staff largely foreign
Open Source SDOF Adds Two Defensive Layers via Intent Router and StateAwareDisp
Open Source D&B rebuilds 642 million‑business database after AI agents hit limits
From Archives Google to roll out AI agents this week, adopting OpenClaw features
From Archives Google AI agents expand on Alerts, moving beyond simple notifications
Common Questions Answered
What is the context window size of Cohere's Command A+ model?
Command A+ features a 128,000-token context window, which allows it to process significantly larger amounts of text in a single request compared to many competing models. This expanded context window enables the model to handle more complex and lengthy documents while maintaining coherence throughout the conversation.
How many languages does Command A+ support?
Command A+ provides support for 48 different languages, demonstrating real multilingual fluency across a diverse range of linguistic contexts. This extensive language support makes it particularly valuable for enterprises operating in multiple regions and languages globally.
What were the performance improvements shown in Command A+ benchmarks?
Command A+ demonstrated transformational performance gains across multiple benchmarks, with agent task scores jumping from 37 to 85 percent on τ²-Bench Telecom and coding test scores leaping from 3 to 25 percent on Terminal-Bench Hard. These substantial improvements represent significant advances rather than incremental gains over its predecessor.
What multimodal capabilities does Command A+ include?
Command A+ has the ability to process both text and images within a single model, enabling it to handle multimodal workflows that enterprises commonly require. This combined text and image processing capability makes it more versatile for complex, real-world applications.
How does Command A+ compare to competing models like Claude 4.5 Haiku and Gemma 4 31B?
Command A+ holds its ground against peers like Claude 4.5 Haiku and Gemma 4 31B without apology, according to the benchmarks presented. As an open-source model with weights available on Hugging Face and ready quantizations, it provides competitive performance while maintaining the accessibility benefits of an open-weight release.
Further Reading
- Introducing Command A+ - Cohere — Cohere Blog
- Command A Plus Model Documentation — Cohere Docs
- Cohere's Command-R+ Specialized Model for RAG and Tools — YouTube
- Enterprise AI Models | Cohere — Cohere
