兔子API 兔子API
中

model

Mistral family

Direct answerMistral currently features Medium 3.5 and Small 4 for general and agentic work, plus dedicated code, OCR, speech, embedding, and safety models with distinct licenses.

Updated · Reviewed

Current official portfolio

Mistral's featured models include Medium 3.5 for multimodal agents and coding, Small 4 combining instruct, reasoning, and coding, and specialized Voxtral, OCR, Shieldstral, and Leanstral offerings.

Open and hosted models

The catalog labels Apache, modified MIT, non-commercial, and Premier licenses. Open weights can be self-hosted while Premier models are services; API access, license, and infrastructure cost are separate selection dimensions.

Choosing within the family

Evaluate Medium for complex multimodal agents and coding and Small for efficiency; use Codestral for completion, OCR for documents, Voxtral for audio, and Shieldstral for moderation. Dedicated models are generally more reliable than simulating the task with general chat.

API use

General chat uses /v1/chat/completions; FIM, embeddings, OCR, audio, and classification have separate resources. This site exposes adapted routes with a Bearer token, and tools or JSON Schema remain model-specific.

Lifecycle

The official catalog maintains a deprecated and retired list. Pin API ID and version and verify license, context, output, deployment, tools, prompt caching, region, and migration deadline.

Use cases

  • General reasoning
  • Coding and OCR
  • Speech multimodality

API protocols

  • /v1/chat/completions

FAQ

Is the model name permanent?

No. Official aliases can move and enabled API IDs can change; pin a tested ID in production and verify availability separately in the model marketplace.

How do I choose between general Mistral models and dedicated Codestral, OCR, or Voxtral resources?

Use general models for broad reasoning and agents and prefer dedicated resources for completion, document parsing, audio, or moderation. Dedicated models are often more reliable than simulating the task with chat, but still compare real quality, latency, and cost.

Do Mistral open weights and hosted APIs share the same license and protocol?

No. The catalog includes open, modified, non-commercial, and Premier licenses; open weights can be self-hosted, while Premier models are services. FIM, OCR, and audio also have separate interfaces, so neither one Chat route nor one license describes the family.

  • Mistral AIA first direct Mistral call uses the Mistral base URL, Authorization: Bearer, and an exact model ID. This site exposes only the adapted /v1/chat/completions route and requires this site's base URL and key. Batch, Guardrails, and Moderation are official Mistral capabilities, not evidence that this route automatically supports or enables them.
  • OpenAI-compatible API guideThis site's OpenAI-compatible routes use a Bearer token; Responses, Chat Completions, and media endpoints have distinct objects and streaming semantics and must match the target model.
  • RAG, embedding, and reranking model guideRAG combines chunking, embeddings, retrieval, reranking, grounded generation, and evaluation; completeness comes from citable evidence and no-answer handling, not a larger model list.

Official sources

  1. Mistral Models Official