LoginSign Up
← Back to all sparks
C

Cohere

INFRA · APIS
Velocity7.5

AI platform providing large language models and NLP tools via API for enterprises.

cohere.com ↗

Cohere is widening from chat into a full enterprise model suite: code, audio, and retrieval.

foundation-modelsmultimodalcode-modelsspeech-to-textmodel-deprecationsenterprise-ai
◆Current state
Cohere is shipping across its whole model lineup: a new Command A+ flagship in May, the North-Mini-Code coding model in June, the Transcribe ASR model earlier in the spring, and Rerank/Embed v4 for retrieval. Alongside the launches, it has been aggressively retiring older Command, Embed, and Aya models plus legacy RAG endpoints. The portfolio is consolidating around the Command A family, embed-v4/rerank-v4, and now code and audio.
◆Where it's heading
Cohere is broadening from a chat-and-retrieval vendor into a multi-modal enterprise model suite, adding speech-to-text and now a code-specialized model, while pruning everything that predates the Command A generation. The steady deprecation cadence signals a deliberate narrowing to a smaller, current set of supported models rather than a sprawling catalog.
◆Prediction
Expect a fast or larger sibling of North-Mini-Code, mirroring the pro/fast split Cohere already ships for Rerank, and continued retirement of pre-Command-A models as customers are steered onto the current generation.

◆Recent moves

  1. 18d ago

    Announcing Cohere’s North-Mini-Code-1.0

    ⚡ SPARK

    Cohere's first code-specialized model, extending a lineup that already spans Command A chat, Rerank/Embed retrieval, and Transcribe audio. Signals a move into developer and coding workloads.

    View source ↗
  2. 1mo ago

    Announcing Cohere’s Command A+

    ⚡ SPARK

    A new flagship in the Command A family, the chat tier the rest of the lineup now consolidates around. Reinforces Command A as the supported baseline as older Command models retire.

    View source ↗
  3. 2mo ago

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

    Hard retirement of Embed v2.0 and the Aya Expanse/Vision 8B models; requests using those IDs now fail. Part of the steady pruning of everything predating the current embed-v4 and Command A generation.

  4. 2mo ago

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

    The detailed version of the same Embed v2.0 / Aya 8B retirement, listing recommended replacements (embed-v3/v4, Command A variants). No additional product change beyond the retirement already noted.

    View source ↗
  5. 3mo ago

    Announcing the Cohere Transcribe model

    ⚡ SPARK

    Cohere's first speech-to-text model, adding an audio-in modality to a lineup that had been text-only. An early marker of the multi-modal expansion the code model now continues.

  6. 3mo ago

    March 26, 2026Announcing the Cohere Transcribe modelWe’re pleased to announce the release of Cohere Transcribe, our first transcription m…

    The detailed Transcribe announcement (languages, API usage, Model Vault pricing) covering the same launch as the prior entry, surfaced twice by the changelog feed.

    View source ↗