Cohere open-sources its strongest model yet
Summary
Cohere has open-sourced its most powerful language model, Command A+, under the Apache 2.0 license. This mixture-of-experts model features 218 billion parameters with 25 billion active and is designed for enterprise applications such as agentic tasks, RAG, and multilingual document processing. Command A+ supports 48 languages, handles both text and images, and boasts a 128,000-token context window. It demonstrates significant performance gains over its predecessor, Command A Reasoning, with scores jumping from 37% to 85% on τ²-Bench Telecom and from 3% to 25% on Terminal-Bench Hard. The model achieves nearly 37 points on the Artificial Analysis Intelligence Index, comparable to Claude 4.5 Haiku, Gemma 4 31B, and Mistral Medium 3.5. Its weights are available on Hugging Face in various quantizations, and it runs on two Nvidia H100 GPUs or a single Blackwell GPU.
Key takeaway
For AI Engineers evaluating large language models for enterprise deployment, Command A+ offers a compelling open-source option. Its 218 billion parameters and 128,000-token context window suit complex agentic tasks and multilingual RAG. You should assess its strong benchmark performance on τ²-Bench Telecom and Terminal-Bench Hard. Consider its compatibility with Nvidia H100 or Blackwell GPUs for your next project.
Key insights
Cohere's Command A+ is a powerful, open-source MoE model designed for enterprise AI, offering strong performance and broad language support.
In practice
- Implement agentic tasks.
- Enhance RAG systems.
- Process multilingual documents.
Topics
- Cohere
- Command A+
- Mixture-of-Experts
- Large Language Models
- Enterprise AI
- RAG Systems
Best for: AI Architect, NLP Engineer, CTO, AI Engineer, Machine Learning Engineer, Director of AI/ML
Related on AIssential
Editorial summary, takeaway, and curation by AIssential. Original article published by The Decoder.