Skip to content

Cohere

cohere.com

Frontier large language models and embeddings for enterprise — Command R+, Embed v4, Rerank v4. Accessible via API, AWS Bedrock, OCI, and Azure.

Last 365 days · 8 changes

Current models (2)

Latest version of each family, grouped by capability. Older versions of each family collapse below.

Model Released Context Input $/Mtok Output $/Mtok Modalities
Command
4K $1 $2 text
Command Nightly
4K $1 $2 text

Model lineup metadata via LiteLLM (MIT). Vendor names are trademarks of their respective owners; pricing and capabilities shown for reference — verify with vendor for binding terms.

Deprecations (2)

  • Retirement of Embed v2.0 and Aya Expanse / Vision 8B

    ⚠ Upgrade required
    • Replace retired embedding models with embed-english-v3.0, embed-multilingual-v3.0 or embed-v4.0.
    • For chat tasks use command-r7b-12-2024, command-a-03-2025, or command-a-reasoning-08-2025 as recommended replacements.
    Breaking changes
    • Retired embedding models: embed-english-v2.0, embed-english-light-v2.0, embed-multilingual-v2.0; requests using these IDs will fail.
    • Retired AI models: c4ai-aya-expanse-8b and c4ai-aya-vision-8b; requests using these IDs will fail.
  • Announcing Major Command Deprecations

    ⚠ Upgrade required
    • Migrate to recommended replacements: command-r-08-2024, command-r-plus-08-2024, or the strongest-performing model command-a-03-2025.
    • Consult the migration guide for alternatives to deprecated models and endpoints.
    Breaking changes
    • Removed models: command-r-03-2024 (alias command-r), command-r-plus-04-2024 (alias command-r-plus), command-light, and command-summarize.
    • Retired fine‑tuning for models command-light, command, command-r, classify, and rerank; existing fine‑tuned models become inaccessible.
    • Deprecated API endpoints: /v1/connectors, /v1/chat parameters connectors & search_queries_only, /v1/generate (legacy generative), /v1/summarize (legacy summarization), /v1/classify, Slack App integration, and the Coral Web UI (chat.cohere.com and coral.cohere.com).

Launches (5)

  • Announcing Cohere's Command A+

    Notable features
    • First Mixture of Experts (MoE) model with 25B active parameters and 218B total parameters
    • Supports vision inputs, reasoning, translation, and agentic tasks in a single model
    • Expanded multilingual support covering 48 languages including all official EU languages
  • Announcing the Cohere Transcribe model

    Notable features
    • Launch of Cohere Transcribe model (cohere-transcribe-03-2026) for audio‑in, text‑out ASR
    • Supports 14 languages: English, German, French, Italian, Spanish, Portuguese, Greek, Dutch, Polish, Vietnamese, Chinese, Arabic, Japanese, Korean
    • Available through Cohere's Audio Transcriptions API with free experimentation tier and paid production Model Vault option
  • Announcing Cohere's Command A Translate Model

    Notable features
    • Supports 23 languages: English, French, Spanish, Italian, German, Portuguese, Japanese, Korean, Chinese, Arabic, Russian, Polish, Turkish, Vietnamese, Dutch, Czech, Indonesian, Ukrainian, Romanian, Greek, Hindi, Hebrew, Persian
    • 111 billion parameters for superior translation quality
    • 16K token context length (8K input + 8K output) enabling longer text handling
  • Announcing Cohere's Command A Reasoning Model

    Notable features
    • 111 billion parameters with 256K context length
    • Support for 23 languages (English and 22 others)
    • Strongest tool use performance in the Command family
  • Announcing Cohere's Command A Vision Model

    Notable features
    • Multimodal support for text + up to 20 images (or 20 MB) per request
    • Enterprise‑focused use cases: chart analysis, table parsing, OCR, scene understanding
    • Official language support: English, Portuguese, Italian, French, German, Spanish

Other updates (1)

  • Cohere's Rerank v4.0 Model is Here!

    Notable features
    • Two model variants: rerank-v4.0-pro (quality) and rerank-v4.0-fast (low latency)
    • Multilingual re‑ranking of English and non‑English documents
    • Support for semi‑structured JSON documents

Beta — feedback welcome: [email protected]