Command A+
AI model · Multimodal model · Text → Text by Cohere
Command A+ is Cohere's most advanced generative model, designed specifically for high-performance enterprise agents and complex reasoning tasks. It is optimized for efficiency, capable of running on a single B200 or two H100 GPUs while maintaining high throughput. The model supports a massive 128k context window and is available through the Cohere API, private cloud deployments, or managed infrastructure like Model Vault.
- Languages
- English, French, Spanish, Italian, German, Portuguese, Japanese, Korean, Chinese, Arabic, Russian
- Context Length
- 128k tokens
- Max Output Tokens
- 64k tokens
- Modality
- Text, Images
- License
- Apache 2.0
- Architecture
- Mixture of Experts (MoE)
Overview
Command A+ is Cohere's most advanced generative model, designed specifically for high-performance enterprise agents and complex reasoning tasks. It is optimized for efficiency, capable of running on a single B200 or two H100 GPUs while maintaining high throughput. The model supports a massive 128k context window and is available through the Cohere API, private cloud deployments, or managed infrastructure like Model Vault.
Core Capabilities
- Agentic Performance: Excels at tool use, agents, and retrieval augmented generation (RAG) tasks.
- Multilingual Support: Supports 48 languages for global communication and discovery.
- Unified Intelligence: Combines vision input support, reasoning, and translation capabilities into single model weights.
- Hardware Efficiency: Fits on 1 x B200 or 2 x H100 GPUs for efficient enterprise deployment.
Vision and Multimodality
- Multimodal Input: Supports both text and image inputs for complex business workflows.
- Visual Reasoning: Capable of analyzing charts, graphs, diagrams, and performing OCR.
- Document Understanding: Interprets complex layouts while retaining spatial relationships.
Pricing
- Input Tokens
- $1.00 per 1M tokens (for existing customers)
- Output Tokens
- $2.00 per 1M tokens (for existing customers)
Specifications
- Context Length
- 128k tokens
- Max Output Tokens
- 64k tokens
- Modality
- Text, Images
- License
- Apache 2.0
- Architecture
- Mixture of Experts (MoE)
Added by AntSpace from public sources (docs.cohere.com).