# Changelog

## 2026-10-22

- Z.ai GLM 5.3 Flash (`zai-glm-5-3-flash`) is now available. (model)
  - A multimodal 320B-parameter (18B active) Mixture-of-Experts model from Z.ai with a 1M-token context window.
  - Built for developer agent workflows, visual coding, and complex tool-use tasks.
  - Pricing: $0.15/M input tokens, $0.03/M cached input, $0.50/M output tokens.

## 2026-10-20

- We released the **Voxtral 3** audio family, our next generation of speech-to-text and text-to-speech models. All four models are in Public Preview. See the [Voxtral 3 announcement](https://mistral.ai/news/voxtral-3/) and the [audio models pricing](/inference/pricing#audio-models). (model)
  - **Voxtral TTS 3** (`voxtral-mini-tts-3-0`): text-to-speech with zero-shot voice cloning, streaming, and contextual delivery. See the [model card](/models/voxtral-mini-tts-3-0).
  - **Voxtral Mini Transcribe 3** (`voxtral-mini-transcribe-3-0`): offline transcription and diarization with timestamps. See the [model card](/models/voxtral-mini-transcribe-3-0).
  - **Voxtral Mini Transcribe Realtime 3** (`voxtral-mini-transcribe-realtime-3-0`): real-time transcription. See the [model card](/models/voxtral-mini-transcribe-realtime-3-0).
  - **Voxtral Small Transcribe 3** (`voxtral-small-transcribe-3-0`): offline transcription and diarization with timestamps. See the [model card](/models/voxtral-small-transcribe-3-0).
- We released **Voice Agents**, a managed real-time bidirectional voice pipeline (microphone capture, STT, LLM, TTS, playback) exposed through the Voice Agents SDK, with telephony, browser, and custom-audio interfaces. See the [Voice Agents overview](/studio/audio/voice-agents/overview) and [Voice Agents pricing](/inference/pricing#voice-agents). (other)

## 2026-10-06

- Mistral Large 4 (`mistral-large-4`) is now available in Public Preview. For details, see the [Mistral Large 4 model card](/models/mistral-large-4-0). (model)
  - Mistral Large 4 is an open-weight, general-purpose multimodal model with a Mixture-of-Experts architecture and a 1M context window.
  - Open weights are coming soon.
  - Launch pricing: 50% off for 2 weeks.

## 2026-09-29

- OCR 4.0 (`mistral-ocr-4-0`) is deprecated and retires on September 30, 2026. Use OCR 4.1 (`mistral-ocr-4-1` or `mistral-ocr-latest`) instead, at the same price. (deprecated)
- Leanstral 1.5 (`labs-leanstral-1-5`), an experimental Labs model, is deprecated and retires on September 30, 2026. (deprecated)
- Z.ai GLM 5.2 (`zai-glm-5-2`) is deprecated and retires on October 31, 2026. Use Z.ai GLM 5.3 (`zai-glm-5-3`) instead, at the same price. (deprecated)

## 2026-09-28

- Z.ai GLM 5.3 (`zai-glm-5-3`) is now Generally Available. (model)

## 2026-08-31

- OCR 4.1 (`mistral-ocr-4-1`) is now Generally Available. (model)

## 2026-07-16

- We released OCR 4.1 (`mistral-ocr-4-1`). `mistral-ocr-latest` and `mistral-ocr-4` now point to it. (model)
- The OCR API `confidence_scores_granularity` parameter now supports `"block"` granularity. `"page"` returns page-level scores only, `"block"` returns page-level and block-level scores, and `"word"` returns page-level and word-level scores. (api)

## 2026-06-30

- We released Leanstral 1.5 (`labs-leanstral-1-5`), an updated Lean 4 formal proof engineering model with improved SFT mixture quality and extended long-context reasoning. This model will be **retired on September 30, 2026**. (model)

## 2026-06-23

- We released OCR 4 (`mistral-ocr-4-0`). `mistral-ocr-latest` now points to it. (model)
  - Introducing `include_blocks` in our OCR API. When set to `true`, each page returns a `blocks` array with paragraph-level bounding boxes and a structural label (`text`, `title`, `list`, `table`, `image`, `equation`, `caption`, `code`, `references`, `aside_text`, `header`, `footer`, `signature`) in reading order. Learn more in our [OCR documentation](/studio/document-processing/basic_ocr#block-extraction). (api)
  - The `pages` parameter in our OCR API now also accepts a string of comma-separated digits and ranges (e.g. `"0,1,2"`, `"0-5"`, or `"0,2-4"`) in addition to a list of integers. (api)

## 2026-05-28

- We launched **Vibe**, our unified agent at [chat.mistral.ai](https://chat.mistral.ai), available in three modes: (other)
  - **[Work](/vibe/work/get-started)** for productivity on web and mobile, with Skills, Workflows, Connectors, Libraries, and scheduled tasks.
  - **[Code](/vibe/code/overview)** for developers, covering the [Vibe CLI](/vibe/code/cli/install-setup), the [VS Code extension](/vibe/code/vs-code-extension/install-authenticate), and [Vibe Code Web](/vibe/code/vibe-code-web/get-started) for remote coding sessions in a managed cloud sandbox.
  - **[Chat](/vibe/chat-legacy/agents)** preserves the legacy Le Chat experience for existing workflows.
- Documentation restructured: previous `/le-chat/*` and `/mistral-vibe/*` paths now redirect to the new `/vibe/*` tree. (other)

## 2026-04-28

- We released Mistral Medium 3.5 (`mistral-medium-3-5`), our frontier-class multimodal model optimized for agentic and coding use cases, with adjustable reasoning via the `reasoning_effort` parameter. Released as open weights under a Modified MIT license. (model)

## 2026-03-23

- We released Voxtral TTS (`voxtral-tts-2603`), our state-of-the-art text-to-speech model with zero-shot voice cloning, multilingual support, and real-time streaming. (model)

## 2026-03-16

- We released Mistral Small 4 (`mistral-small-2603`), a hybrid model unifying instruct, reasoning, and coding in a single multimodal model with a 256k context window. (model)
- We released Leanstral (`labs-leanstral-2603`), our first open-source code agent designed for Lean 4 formal proof engineering. (model)

## 2026-03-12

- We released Mistral Moderation 2603 (`mistral-moderation-2603`). (model)
- We added [Custom Guardrails](/studio/safety-moderation#custom-guardrails) support for Agents and Conversations. (api)
    - Guardrails can now be configured directly 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.

## 2026-02-04

- We released Voxtral Mini Transcribe 2 (`voxtral-mini-2602`) and Voxtral Mini Transcribe Realtime (`voxtral-mini-transcribe-realtime-2602`). (model)
    - Introducing `context biasing` in our Audio Transcriptions API,
    - Introducing `diarize`in our Audio Transcriptions API.

## 2026-01-27

- We released Vibe 2.0. (other)
    - Devstral 2.0 moves to paid API access. (model)
- Document Annotations update and improvements, 8 page limit removed. (api)

## 2026-01-19

- We released inline batching allowing the creation of batch jobs without file uploading. (api)

## 2025-12-18

- We released OCR 3 (`mistral-ocr-2512`). (model)
  - Introducing `table_format` in our OCR API, allowing you to choose between `markdown` and `html` for table formatting. (api)
  - Introducing `extract_footer`, `extract_header` in our OCR API, as well as `hyperlinks` output. (api)

## 2025-12-16

- We released Mistral Small Creative (`labs-mistral-small-creative`) as a Labs model. (model)

## 2025-12-09

- We released Devstral 2 (`devstral-2512`) and Devstral Small 2 (`labs-devstral-small-2512`). (model)
- We released Mistral Vibe. (other)

## 2025-12-02

- We released Mistral Large 3 (`mistral-large-2512`) and Ministral 3 (`ministral-3b-2512`, `ministral-8b-2512` and `ministral-14b-2512`). (model)

## 2025-09-17

- We released Magistral Medium 1.2 (`magistral-medium-2509`) and Magistral Small 1.2 (`magistral-small-2509`). (model)

## 2025-08-27

- Added a new parameter `p` to the chunks streamed back by the Completion API. (security) (api)
  - Implemented for security to prevent token-length side-channel attacks, as reported by Microsoft researchers.
  - Note that this change may break applications relying on strict parsing of the chunks. Applications using the official SDK are unaffected, but users relying on the `mistral-common` package may need to update to `1.8.4` or higher.

## 2025-08-12

- We released Mistral Medium 3.1 (mistral-medium-2508). (model)

## 2025-07-30

- We released Codestral 2508 (`codestral-2508`). (model)

## 2025-07-24

- We released Magistral Medium 1.1 (`magistral-medium-2507`) and Magistral Small 1.1 (`magistral-small-2507`). (model)
- We released a Document Library API to manage libraries. (api)
- SDK support for Audio and Transcription available. (other)

## 2025-07-15

- We released our first Audio models for chat and a Transcription API:
  - Voxtral Small (`voxtral-small-2507`) available for chat use cases (model)
  - Voxtral Mini (`voxtral-mini-2507`) available for chat use cases (model)
  - Voxtral Mini Transcribe (`voxtral-mini-2507` via `audio/transcriptions`) optimized for transcription (model) (api)

## 2025-07-10

- We released Devstral Small 1.1 (`devstral-small-2507`) and Devstral Medium (`devstral-medium-2507`). (model)

## 2025-06-23

- Mistral Small 3.2 API available (`mistral-small-2506`). (api)

## 2025-06-20

- We released Mistral Small 3.2. (model)

## 2025-06-10

- We released Magistral Medium (`magistral-medium-2506`) and Magistral Small (`magistral-small-2506`). (model)

## 2025-05-28

- We released Codestral Embed (`codestral-embed`). (model)

## 2025-05-27

- We released the new [Agents API](/studio/agents/introduction). (api)

## 2025-05-22

- We released Mistral OCR 2 (`mistral-ocr-2505`) and [annotations](/studio/document-processing/annotations). (model) (api)

## 2025-05-21

- We released Devstral Small (`devstral-small-2505`). (model)

## 2025-05-07

- We released Mistral Medium 3 (`mistral-medium-2505`). (model)

## 2025-04-16

- We released the [Classifier Factory](/resources/deprecated/finetuning/classifier_factory). (api)

## 2025-03-17

- We released Mistral Small 3.1 (`mistral-small-2503`). (model)

## 2025-03-06

- We released Mistral OCR (`mistral-ocr-2503`) and [document understanding](/studio/document-processing/basic_ocr). (model) (api)

## 2025-02-17

- We released Mistral Saba (`mistral-saba-2502`). (model)

## 2025-01-30

- We released Mistral Small 3 (`mistral-small-2501`). (model)

## 2025-01-28

- We released custom [structured outputs](/studio/conversations/structured-output#custom-structured-outputs) for all models. (api)

## 2025-01-13

- We released Codestral 2501 (`codestral-2501`). (model)

## 2024-11-18

- We released Mistral Large 2.1 (`mistral-large-2411`) and Pixtral Large (`pixtral-large-2411`). (model)
- [Le Chat](https://chat.mistral.ai/): (other)
  - Web search with citations
  - Canvas for ideation, in-line editing, and export
  - State of the art document and image understanding, powered by the new multimodal Pixtral Large
  - Image generation, powered by Black Forest Labs Flux Pro
  - Fully integrated offering, from models to outputs
  - Faster responses powered by speculative editing

## 2024-11-06

- We released moderation API and batch API. (model) (api)
- We introduced three new parameters: (api)
  - `presence_penalty`: penalizes the repetition of words or phrases
  - `frequency_penalty`: penalizes the repetition of words based on their frequency in the generated text
  - `n`: number of completions to return for each request, input tokens are only billed once.

- We downscaled the temperature parameter of `pixtral-12b`, `ministral-3b-2410`, and `ministral-8b-2410` by a multiplier of 0.43 to improve consistency, quality, and unify model behavior. (api)

## 2024-10-09

- We released Ministral 3B (`ministral-3b-2410`) and Ministral 8B (`ministral-8b-2410`). (model)

## 2024-09-17

- We released Pixtral (`pixtral-12b-2409`) and Mistral Small v24.09 (`mistral-small-2409`). (model)
- We reduced price on our flagship model, Mistral Large 2. (api)
- We introduced a free API tier on La Plateforme. (api)

## 2024-09-13

- In le Chat, we added a mitigation against an obfuscated prompt method that could lead to data exfiltration, reported by researchers [Xiaohan Fu](https://xhfu.me/) and Earlence Fernandes. The attack required users to willingfully copy and paste adversarial prompts and provide personal data to the model. No user was impacted and no data was exfiltrated. (security)

## 2024-07-29

- We released version 1.0 of our Python and JS SDKs with major upgrades and syntax changes. Check out our [migration guide](https://github.com/mistralai/client-python/blob/main/MIGRATION.md) for details. (other)
- We released Agents API. See details [here](/studio/agents/introduction). (api)

## 2024-07-24

- We released Mistral Large 2 (`mistral-large-2407`). (model)
- We added fine-tuning support for Codestral, Mistral Nemo and Mistral Large. Now the model choices for fine-tuning are `open-mistral-7b` (v0.3), `mistral-small-latest` (`mistral-small-2402`), `codestral-latest` (`codestral-2405`), `open-mistral-nemo` and , `mistral-large-latest` (`mistral-large-2407`) (api)

## 2024-07-18

- We released Mistral Nemo (`open-mistral-nemo`). (model)

## 2024-07-16

- We released Codestral Mamba (`open-codestral-mamba`) and Mathstral. (model)

## 2024-06-05

- We released fine-tuning API. Check out the [capability docs](/resources/deprecated/finetuning) and [guides](/resources/deprecated/finetuning). (api)

## 2024-05-29

- New model available: `codestral-latest` (aka `codestral-2405`). Check out the code generation [docs](/vibe/code/overview). (model) (api)

## 2024-05-23

- Function calling: `tool_call_id` is now mandatory in chat messages with the `tool` role. (api)

## 2024-04-17

- New model available: `open-mixtral-8x22b` (aka `open-mixtral-8x22b-2404`). Check the release [blog](https://mistral.ai/news/mixtral-8x22b/) for details. (model)
    - For function calling, `tool_call_id` must not be null for `open-mixtral-8x22b`. (other)
- We released three versions of tokenizers for commercial and open-weight models: check the related [guide](https://github.com/mistralai/mistral-common) and [repo](https://github.com/mistralai/mistral-common) for more details. (other)

## 2024-03-28

- JSON mode now available for all models on La Plateforme. (api)

## 2024-02-26

- API endpoints: We renamed 3 API endpoints and added 2 model endpoints. (api)

  - `open-mistral-7b` (aka `mistral-tiny-2312`): renamed from `mistral-tiny`. The endpoint `mistral-tiny` will be deprecated in three months.
  - `open-mixtral-8x7B` (aka `mistral-small-2312`): renamed from `mistral-small`. The endpoint `mistral-small` will be deprecated in three months.
  - `mistral-small-latest` (aka `mistral-small-2402`): new model. (model)
  - `mistral-medium-latest` (aka `mistral-medium-2312`): old model. The previous `mistral-medium` has been dated and tagged as `mistral-medium-2312`. The endpoint `mistral-medium` will be deprecated in three months.
  - `mistral-large-latest` (aka `mistral-large-2402`): our new flagship model with leading performance. (model)

- New API capabilities: (api)

  - [Function calling](/studio/conversations/function-calling): available for Mistral Small and Mistral Large.
  - [JSON mode](/studio/conversations/structured-output#json-mode): available for Mistral Small and Mistral Large

- [La Plateforme](https://console.mistral.ai/): (other)

  - We added multiple currency support to the payment system, including the option to pay in US dollars.
  - We introduced enterprise platform features including admin management, which allows users to manage individuals from your organization.

- [Le Chat](https://chat.mistral.ai/): (other)
  - We introduced the brand new chat interface Le Chat to easily interact with Mistral models.
  - You can currently interact with three models: Mistral Large, Mistral Next, and Mistral Small.

## 2024-01-16

- We added token usage information in streaming requests. You can find it in the last chunk returned. (api)

## 2024-01-11

- We have enhanced the API's strictness. Previously the API would silently ignores unsupported parameters in the requests, but it now strictly enforces the validity of all parameters. If you have unsupported parameters in your request, you will see the error message "Extra inputs are not permitted". (api)
- A previous version of the [guardrailing documentation](/studio/safety-moderation) incorrectly referred to the API parameter as `safe_mode` instead of `safe_prompt`. We corrected this in the documentation. (other)