Cohere
last releaseAugust 21, 2025
powered byCommand R+, Embed 4
goblin vibe check:
the api you reach for when you are building a search or rag product and need reliable embeddings and retrieval without paying openai prices
enterprise-focused llm platform offering command models for text generation, embed models for semantic search, and rerank for retrieval pipelines. less consumer-facing than openai or anthropic, but strong for b2b rag pipelines, search, and classification at scale. api pricing is competitive and the retrieval-focused tooling is a genuine differentiator. not the right choice for general chatbot use or creative tasks.
context
128k
tokens
cost
$3
per 1M in tokens (Command R+)
Command A is built for tool use, RAG, agents, and multilingual enterprise workloadsEmbed 4 adds multimodal retrieval with unified text and image embeddings256K context on Command A and 128K context on Embed 4 support larger corporaCohere stays strongest in enterprise retrieval and secure internal search
key features
Command A is built for tool use, RAG, agents, and multilingual enterprise workloadsEmbed 4 adds multimodal retrieval with unified text and image embeddings256K context on Command A and 128K context on Embed 4 support larger corporaCohere stays strongest in enterprise retrieval and secure internal search
spec & usage
Command A Reasoning launched on Aug 21, 2025
Command A only needs two A100 or H100-class gpus for self-hosted inference
scope:
datalanguagesearchautomationwritingapicloudpaidbenchmark-strongfine-tunable
last releaseAugust 21, 2025