←── back to feed
/topics/cohere-releases-command-a-sparse-moe-model

Cohere releases Command A+ sparse MoE model

2 items2 sourcesupdated 26d agotrend 0

Cohere released Command A+, a 218-billion-parameter sparse mixture-of-experts model designed for agentic workflows that consolidates four prior Command A variants. The model runs on as few as two H100 GPUs at W4A4 quantization, supports 48 languages, and is Cohere's first multimodal reasoning model.

  • 218B parameters in sparse MoE architecture
  • Runs on minimum two H100 GPUs with W4A4 quantization
  • Consolidates four prior Command A variants into single model
  • Supports 48 languages
  • First Cohere multimodal reasoning model