AI news story
Cohere Releases Command A+: A 218B Sparse MoE Model for Agentic Workflows That Runs on as Few as Two H100 GPUs
Cohere releases Command A+, an open-source 218B Sparse Mixture-of-Experts model consolidating four prior Command A variants i…
Editor's take
Cohere has unveiled Command A+, a large, open-source language model designed for agentic workflows. This 218 billion parameter Sparse Mixture-of-Experts model integrates previous Command A iterations and can operate efficiently, requiring as few as two NVIDIA H100 GPUs with W4A4 quantization.
The significance lies in its accessibility and specific design for complex, multi-step AI agent tasks. By consolidating models and optimizing for hardware, Cohere lowers the barrier to entry for deploying sophisticated AI agents, potentially accelerating development in fields like automated customer service, complex data analysis, and personalized digital assistants. Its 48-language support and multimodal capabilities further broaden its applicability.
Future developments to monitor include the real-world performance of Command A+ in diverse agentic applications, particularly its inference speed and cost-effectiveness compared to other leading models like OpenAI's GPT-4. The adoption rate by developers and enterprises, and whether this model truly democratizes advanced agent development, will be key indicators of its impact.