Every atomic assertion extracted from the underlying record, ranked by evidence strength.
`mistral-medium-3-5` is a multimodal model.
`mistral-medium-3-5` is optimized for agentic use cases.
`mistral-medium-3-5` is optimized for coding use cases.
`mistral-medium-3-5` supports adjustable reasoning via the `reasoning_effort` parameter.
`mistral-medium-3-5` is released as open weights.
`mistral-medium-3-5` is released under a Modified MIT license.
Mistral AI released `voxtral-tts-2603`.
`voxtral-tts-2603` is a text-to-speech model.
`voxtral-tts-2603` supports zero-shot voice cloning.
`voxtral-tts-2603` supports multilingual capabilities.
`voxtral-tts-2603` supports real-time streaming.
Mistral AI released `mistral-small-2603`.
`mistral-small-2603` is a hybrid model.
`mistral-small-2603` supports instruct capabilities.
`mistral-small-2603` supports reasoning capabilities.
`mistral-small-2603` supports coding capabilities.
`mistral-small-2603` is a multimodal model.
`mistral-small-2603` has a 256k context window.
Mistral AI released `labs-leanstral-2603`.
`labs-leanstral-2603` is an open-source code agent.
`labs-leanstral-2603` is designed for Lean 4 formal proof engineering.
Mistral AI released `mistral-moderation-2603`.
`mistral-moderation-2603` is a moderation model.
Mistral AI added Custom Guardrails support for Agents.
Mistral AI added Custom Guardrails support for Conversations.
Guardrails can be configured on an Agent via the `guardrails` parameter.
Guardrails can be passed per-request on `POST /v1/conversations` using the `guardrails` field.
Guardrails can be passed per-request on `POST /v1/chat/completions` using the `guardrails` field.
Mistral AI released `voxtral-mini-2602`.
Mistral AI released `voxtral-mini-transcribe-realtime-2602`.
Mistral AI introduced context biasing in its Audio Transcriptions API.
Mistral AI introduced diarization in its Audio Transcriptions API.
Devstral 2.0 moved to paid API access.
Mistral AI updated and improved Document Annotations.
The 8-page limit for Document Annotations was removed.
Mistral AI released inline batching.
Inline batching allows creation of batch jobs without file uploading.
Mistral AI released `mistral-ocr-2512`.
Mistral AI introduced `table_format` in its OCR API.
The `table_format` parameter in the OCR API allows choosing between markdown and html for table formatting.
Mistral AI introduced `extract_footer` in its OCR API.
Mistral AI introduced `extract_header` in its OCR API.
Mistral AI introduced hyperlinks output in its OCR API.
Mistral AI released `labs-mistral-small-creative`.
`labs-mistral-small-creative` is a Labs model.
Mistral AI released `devstral-2512`.
Mistral AI released `labs-devstral-small-2512`.
Mistral AI released Mistral Vibe.
Mistral AI released `mistral-large-2512`.
Mistral AI released `ministral-3b-2512`.
Mistral AI released `ministral-8b-2512`.
Mistral AI released `ministral-14b-2512`.
Mistral AI released `magistral-medium-2509`.
Mistral AI released `magistral-small-2509`.
Mistral AI added a new parameter `p` to chunks streamed by the Completion API.
The new parameter `p` in the Completion API was implemented for security.
The new parameter `p` in the Completion API prevents token-length side-channel attacks.
The change to the Completion API may break applications relying on strict parsing of chunks.
Applications using the official SDK are unaffected by the Completion API change.
Mistral AI released `mistral-medium-3-5`.