Command A

Cohere, a Canadian AI startup, is excited to introduce Command A, a high-performance AI model perfect for businesses. It''s fast, efficient, and supports many languages, making it a super useful tool for companies.

Key Features

Command A shines in efficiency and performance. It can run on just two GPUs, which is a big step up from other models that need up to 32 GPUs. It processes 156 tokens per second, making it much quicker than similar models. With a context length of 256,000 tokens, it can handle large inputs equal to about 600 pages of text. This makes it ideal for tasks that need lots of information.

Command A supports 23 global languages, including improved handling of Arabic dialects. It scored 98.2% accuracy in responding to English prompts in Arabic, which is better than other models. This makes it very useful for businesses that work in many languages.

Benefits

Command A brings several benefits for enterprises. It has advanced retrieval-augmented generation, which means it can give highly accurate responses. It also supports complex workflows by working well with enterprise tools. Plus, it integrates with Cohere''s North AI platform, allowing businesses to automate tasks using secure, enterprise-grade AI agents. Private deployments are up to 50% cheaper than API-based access, making it a cost-effective solution.

Use Cases

Command A is designed for enterprise applications. Its multilingual capabilities make it useful for global businesses. Its advanced features allow it to handle complex workflows and integrate with enterprise tools, making it a versatile tool for various business needs.

Cost Price

Command A is available on the Cohere platform and with open weights for research use only on Hugging Face under a Creative Commons Attribution Non Commercial 4.0 International (CC-by-NC 4.0) license. The cost is $2.50 per million input tokens and $10.00 per million output tokens. Private and on-prem deployments are available upon request.

Reviews Testimonials

Several AI researchers and Cohere team members have shared their excitement for Command A. Dwaraknath Ganesan, pretraining at Cohere, commented that Command A is amazing and can be deployed on just 2 H100 GPUs. Pierre Richemond, AI researcher at Cohere, added that Command A is a new model optimized for efficiency in enterprise use cases.