Cohere released Command A Reasoning, its first reasoning large language model designed specifically for enterprise use cases like customer service, market research, and data analysis. The 111-billion parameter model supports 256,000 tokens, 23 languages, and features a token budget system that allows developers to control reasoning depth versus speed. It outperforms competitors like DeepSeek-R1 and Mistral on enterprise benchmarks and includes built-in tool use capabilities for API integration. The model is available for research on Hugging Face and commercially through Cohere's platform with custom enterprise pricing.

6m read timeFrom venturebeat.com
Post cover image
Table of contents
Controlled thinkingExcels at enterprise targeted benchmarksA focus on tool use and agentic workflowsSafety: built for high-stakes workEarly buy-in from large enterprisesAvailability and licensing

Sort: