Cohere

Cohere is an AI company focused on developing enterprise-grade language models, offering multiple series of high-performance models including the Command series and Aya Expanse series.

Major Model Series:

Command Series:

• Command-R+: One of the highest quality models, supporting 128k context window
• Command-R: Balanced performance and pricing, supporting 128k context window
• Command-R+ (Apr ‘24): Latest version with performance optimizations
• Command-R (Mar ‘24): Speed-optimized version achieving 163 tokens/second

Aya Expanse Series:

• Aya Expanse 8B: Lightweight high-speed model with 8k context window
• Aya Expanse 32B: Large-scale high-performance model with 8k context window

Performance Metrics: Speed Performance:

• Command-R (Mar ‘24): 163 tokens/second
• Aya Expanse 8B: 139 tokens/second
• Other models maintain high output speeds

Latency Performance:

• Aya Expanse 32B: 0.21 seconds
• Command-R (Mar ‘24): 0.21 seconds
• Generally low latency across all models

Pricing Strategy (per million tokens):

• Command-R: $0.26
• Command-R (Mar ‘24): $0.75
• Command-R+: $4.38
• Command-R+ (Apr ‘24): $6.00
• Aya Expanse series: $0.75

Technical Features:

• Function calling support across all models
• Complete JSON mode output support
• Stable and reliable API services
• Enterprise-grade security and scalability

Performance Evaluation:

• Excellent results in MMLU (reasoning and knowledge) tests
• High scores in GPQA (scientific reasoning) evaluation
• Outstanding capabilities in mathematical reasoning and code generation
• Strong showing in LMSys Chatbot Arena ELO ratings

Cohere provides comprehensive AI solutions for enterprises and developers through diverse model offerings and flexible pricing strategies. The company particularly emphasizes model practicality and reliability, committed to delivering high-quality language model services to users.