Mistral Release Notes

Follow

124 release notes curated from 61 sources by the Releasebot Team. Last updated: Nov 6, 2026

Get this feed:

Mistral Products

  • November 2026
    • No date parsed from source.
    • First seen by Releasebot:
      Nov 6, 2026
    Mistral logo

    Mistral

    Le Chonk brings the brains. Studio puts it to work.

    Mistral expands its enterprise AI platform with Mistral Large 4 and new Studio capabilities for managed agents, workflows and apps. Teams can build and deploy AI solutions across cloud, VPC and on-premises environments, with greater control, observability and evaluation tools.

    Mistral Large 4 (Le Chonk) enhances enterprise AI with advanced coding, cybersecurity, and document reasoning, while Mistral Studio provides a flexible platform for deploying agents, workflows, and apps across cloud, VPC, or on-premises environments. New features like managed agents, workflows, and enterprise context streamline business processes, enabling faster development and deployment of AI solutions. Customers like ABANCA and BNP Paribas already use Studio to automate tasks, improve efficiency, and maintain full control over their AI systems.

    Mistral Large 4 (Le Chonk) pushes the frontier of what enterprises can build with open models. It does great for coding, delivers advanced defensive cybersecurity capabilities, and brings powerful reasoning to complex documents and business tasks.

    Mistral Studio is our enterprise AI platform for putting that intelligence to work in your business. It lets your teams activate models quickly through agents, workflows and apps connected to your enterprise data and systems. Deploy them wherever you choose: in the cloud, in your own VPC, on-premises or across a combination of environments. And retain full control and understanding of the intelligence you create, from how your systems work to the knowledge they build.

    Today, we’re expanding Studio with managed agents and apps, alongside new capabilities for workflows, enterprise context, agent observability and evaluations.

    Customers and partners including ABANCA, Accenture, Alten, BNP Paribas, Capgemini, CMA CGM and Tata Consultancy Services already use Studio in production. With this release, your developers can build and deploy new use cases faster, then reuse that work across more of the business.

    Agents, workflows and apps for your business processes

    Agents, workflows and apps serve different parts of a business process. An agent investigates and acts, a workflow coordinates execution and approvals, and an app gives employees a way to use these systems.

    Studio brings all three together on shared infrastructure. You start with a model like Le Chonk and connect it to your enterprise data and systems through connectors and MCP servers. That foundation lets you easily build agents, workflows and apps with access to the context they need. Studio manages how they run and lets teams inspect what happened.

    Through the new Mistral MCP server, developers can build with Mistral Vibe or the coding agent they already use, test locally and deploy to their Studio environment with one command. Studio handles hosting and execution, so teams can build on that foundation to get each new use case into production faster.

    Build managed agents that investigate and act

    In Studio, developers can now build agents that carry out complex, multi-step tasks. Developers can equip them with skills that capture the team’s procedures, sandboxes where they can run code, and subagents to delegate parts of a task. Hooks let developers add custom logic at key points in execution, shaping how the agent works and integrates with existing systems.

    Studio manages the runtime behind all that work, using the proven infrastructure that supports millions of Mistral Vibe sessions each week. Developers can focus on the agent’s capabilities and behaviour, while setting boundaries on the tools, data and actions it can access. That means getting capable agents into production faster, with less infrastructure to build and more control over how your agents operate.

    Our engineering team built an agent in Studio to investigate production incidents. When an alert fires, engineers need to work out what happened, often under pressure to resolve the issue. The agent starts the investigation: it follows the team’s runbook, asks for approval before taking sensitive actions and posts its findings in Slack. Engineers pick up the investigation with useful context already gathered, saving the team hundreds of hours each month.

    Automate processes with workflows

    Workflows define how a process runs: which steps follow fixed rules, where an agent reasons and when a person decides. Interrupted runs can resume where they left off. Teams can describe a process in Mistral Vibe and use it to generate, test and deploy a workflow, with code they can inspect and edit. Today’s release also adds event triggers from enterprise systems and a visual graph that makes each workflow’s logic easier to inspect.

    Our product operations team built a workflow in Studio to sort ideas from internal teams and feedback from customers. Routing each submission to the right people is one of the team’s routine tasks, but one that takes time. A model identifies bugs and feature requests, then routing rules send them to the appropriate team. This saves the product operations team about 35 hours a week.

    Build apps ready for enterprise use

    Coding agents make it easy to build an application, but putting it to work across an enterprise means meeting requirements for security, access and reliability. Studio provides hosting, authentication and access controls, helping developers get applications ready for enterprise use faster. Teams can build with pre-approved, reusable components, reducing the amount of new code that needs security review.

    These applications can use agents and workflows, retrieve enterprise documents through search indexes and access shared context in Studio. Developers can test locally, then deploy with one command to make the application available at its own link or inside Mistral Vibe.

    Our Applied AI team, which helps customers build AI solutions, built a staffing app in Studio. Previously, staffing a project meant manually matching team profiles and past experience against requirements spread across documents and tools.

    The app uses an agent to recommend people and explain how their experience fits the project. Staffing leads get a shortlist with supporting evidence ready to review, helping Mistral assemble teams faster and uncover expertise that might otherwise be overlooked when staffing complex customer AI projects.

    Each use case leaves the enterprise better equipped to build the next. More engineering effort goes into improving operations, with less time spent rebuilding the infrastructure around them.

    Keep your AI sovereign

    Studio gives you control over where your AI runs, how it operates and the intelligence you build. Deploy both the models and the platform in your own cloud, VPC or on-premises environment. You can also use Mistral-hosted deployments or combine environments to meet your requirements. Sensitive data and the processes built around it can stay within your perimeter.

    You choose the models, govern access to data and tools, and decide which actions require human approval. OpenTelemetry traces show each step of a run, including model calls, tool use and approvals, so your teams can see where a process slows down or fails. Evaluations let them test changes to prompts, models or tools against your own datasets before release. Those traces and evaluations remain theirs, preserving the operational knowledge they build through use.

    Customers are already putting this into practice with systems built alongside our Solutions team. From penetration testing to KYC onboarding, they’re bringing AI into critical operations while keeping sensitive data and specialist knowledge under their control.

    Get started

    Choose a process your team wants to improve and put Le Chonk to work in Studio.

    • Build an agent: Managed Agents documentation · Incident investigation cookbook
    • Automate a process: Workflows documentation · Workflow cookbooks
    • Create an app: Apps documentation · Document review cookbook

    Explore interactive demos and enterprise use cases, or read the release notes for the full list of updates.

    For an enterprise rollout or a business-critical implementation, talk to our team about what you want to automate and where it needs to run.

    Original source
  • Nov 2, 2026
    • Date parsed from source:
      Nov 2, 2026
    • First seen by Releasebot:
      Nov 3, 2026
    Mistral logo

    Mistral Common by Mistral

    v1.13.1: Hotfix to allow developer role

    Mistral Common releases updates that map OpenAI developer roles to system and add shared timeouts for image and audio downloads.

    What's Changed

    • Remap OpenAI "developer" role to "system" in from_openai by @juliendenize in #430
    • Move v15 instruct tokenizer tests into the unit suite with model settings and JSON schema coverage by @juliendenize in #418
    • Add shared download timeout for image and audio downloads by @juliendenize in #432
    • Version 1.13.1 by @juliendenize in #437

    Full Changelog: v1.13.0...v1.13.1

    Original source
  • All of your release notes in one feed

    Join Releasebot and get updates from Mistral and hundreds of other software products.

    Create account
  • November 2026
    • No date parsed from source.
    • First seen by Releasebot:
      Nov 2, 2026
    Mistral logo

    Mistral

    MISTRAL LARGE 4 (LE CHONK)

    Mistral releases improved Mistral Large 4 open weights, advancing agentic coding and defensive cybersecurity with stronger reasoning efficiency and lower token use. The multimodal model adds three reasoning modes and is available to try, customize, and deploy on users’ own infrastructure.

    Built for Coders, Trusted by Defenders

    We release the weights of Le Chonk, advancing the frontier of open models in agentic coding and cybersecurity.

    Earlier this month, we released Mistral Large 4 (Le Chonk), our flagship 1-trillion-parameter, natively multimodal model, in public preview. The response to the preview has exceeded our expectations, and we're grateful to our developer community, early-access customers, and partners for their support, collaboration, and feedback.

    Since then, we've put that feedback to work while continuing to scale reinforcement learning, driving massive improvements in coding and defensive cybersecurity. Gains continue to accelerate, with no signs of saturation.

    Today, we're releasing the weights of this improved Mistral Large 4. The new model also features substantial gains in reasoning efficiency that give developers greater performance while lowering inference costs.

    Mistral Large 4 ranks as the best open-weight model for coding and among the world's most capable open models overall. Trained from scratch on European infrastructure with clear lineage from inception to release, it gives organizations a trusted foundation for Sovereign AI.

    Mistral is defining the sovereign AI category and leading the open frontier. We're committed to building the world's most capable models and releasing them openly, giving developers, enterprises, and institutions the power to own, customize, and control their AI.

    Built for the Builders

    Mistral Large 4 entered public preview with strong results across leading coding benchmarks. Just three weeks of additional reinforcement learning have since propelled it to the forefront of open-weight coding, taking the lead on FrontierCode and Terminal-Bench while matching the best open-weight models on the Artificial Analysis Coding Agent Index with substantially greater token efficiency.

    On Cognition's private FrontierCode 2.0 benchmark, which assesses end-to-end code quality and whether a repository maintainer would merge the agent's change, Mistral Large 4 leads with a score of 19.7%.

    Pushing the pareto frontier for agentic coding

    Mistral Large 4 delivers frontier coding performance with significantly fewer tokens, making sophisticated coding agents more economical to run at scale.

    On the Artificial Analysis Coding Agent Index, Mistral Large 4 scores 48%, effectively tied with the leading open-weight model, GLM-5.3, while using 47% fewer tokens per task on average.

    The gains extend to terminal-based coding workflows. On Terminal-Bench 4.0, Mistral Large 4 improved by 13 percentage points to 41% compared to its public preview, taking the lead among open-weight models while reducing output tokens by 30%.

    To give developers greater control over this tradeoff between performance and cost, Mistral Large 4 now supports three reasoning modes:

    Mode Recommended Use Low Simple tasks; replaces ‘None’ reasoning mode Medium Everyday tasks High (default) Hardest use cases requiring high precision and thorough reasoning

    Beyond benchmarks, Mistral Large 4's coding capabilities are validated by human preference. In Surge Code's blind pairwise evaluations, Mistral Large 4 was preferred over leading open-weight models including GLM-5.3, Kimi K3, and DeepSeek V4.1 Flash, ranking behind only Claude Opus 5.5 among the models evaluated.

    Built for the Defenders

    Since its preview, we've been working with leading cybersecurity vendors to evaluate Mistral Large 4 in real-world environments to test Mistral Large 4’s strengths in vulnerability reproduction and patching.

    Mistral Large 4 scores 51% on the Artificial Analysis Cyber Index, which measures a model’s ability to discover vulnerabilities in a codebase, reproduce and validate them, and patch them without breaking existing functionality. Crucially, Mistral Large 4 handles these legitimate defensive security tasks without unnecessary refusals.

    Safety has improved alongside capability. Our model’s built-in guardrails distinguishes between defensive and malicious, offensive cybersecurity tasks. It retains strong performance on Cyber benchmarks, while achieving an average refusal rate of 96.6% on malicious cyber requests, the highest we measure among open models.

    Here’s what global leaders in cybersecurity have to say about Mistral Large 4:

    In the time since the preview, we carefully monitored cybersecurity-related use in our live traffic. Though we observed the model refusing to comply with black-box attacks against live targets, we see a potential for uplift to conduct real attacks. We are modifying our guardrails accordingly, and will continue updating them in the months to come.

    We are also providing enterprise design partners such as Orange Cyberdefense, with preview access to a specialized variant of Mistral Large 4 optimized for advanced cybersecurity workflows, including penetration testing, vulnerability assessment, and red-teaming. Access is subject to rigorous vetting and contractual safeguards, with testing restricted to owned or explicitly authorized systems in controlled environments.

    Meanwhile, customers are tailoring Mistral Large 4 with proprietary security data, policies, detection logic, and threat environments using Mistral Forge. For well-defined cybersecurity tasks, it can be further optimized into smaller, task-specific variants for greater accuracy, making security applications more effective.

    Built for Work

    Mistral Large 4 has continued improving on a range of broad enterprise-use metrics.

    For finance-related work, we improved the model’s ability to perform real-world workflows, especially ones involving the creation and editing of spreadsheets for finance and accounting use cases. This is reflected in our Finch scores climbing to 73.3%, second only to Kimi K3 for open models.

    Web browsing is a critical part of a majority of knowledge work. Our RL pipeline enabled us to reach 84.5% on BrowseComp (starting from an SFT checkpoint that was achieving 75%). This brings Mistral Large 4 at a higher performance than models like GLM-5.3 or DeepSeek Flash 4.1 1017.

    Agents deployed by an organization need to be secure. Whether deployed to internal members or external users, agents shouldn’t open you up to threats like system prompt exfiltration or malicious code injection. RL increased our gains on the B3 Agent Security Benchmark, with Mistral Large 4 resisting 97.1% of attacks vs 93.3% in Preview. This places it at the top of the OSS frontier, meaning anyone hosting the model can run a safe agent without additional tuning.

    See it in Action

    • Cyber Vulnerability Triage: an agent reads incoming vulnerability reports, investigates the source code and product context, reproduces the flaws in an isolated sandbox, then writes and iterates on a patch until a canary exploit no longer works.
    • Agentic webdev: with no human help, Mistral Large 4 builds Orbital Trace, a live 3D map of roughly 19,000 satellites and debris pieces orbiting Earth. It pulls real tracking data, writes the app, uses it in a browser to catch its own bugs and slowdowns, then fixes what an independent reviewer flags until all requirements pass.
    • Quantum chemistry solver: Mistral Large 4 writes, from scratch, software that calculates where electrons sit in a molecule, and checks its answers against a standard chemistry tool. It then builds a 3D page around it: pick a molecule such as benzene or water, drag a slider to pull it apart, and watch the electrons regroup as the bond breaks.

    Built for Sovereignty

    As organizations build AI around their knowledge and workflows, sovereignty means that they retain control of the models they depend on and the intelligence they create.

    With Mistral Large 4, that control begins with the model itself. Organizations have the ability to inspect, customize, and deploy the model on their own infrastructure. They can use and improve Mistral Large 4 without depending on Mistral to serve it through an API. Mistral Large 4 was developed and trained from scratch in Europe, giving organizations clarity about the origins of the technology they’re building on.

    Control also extends to how AI operates inside an organization. Organizations can customize Mistral Large 4 around their proprietary data with Mistral Forge, and they can build, govern, and improve the resulting agents and applications using Mistral Studio. Mistral’s Applied AI experts can work alongside customer teams to build and deploy these tailored systems. Through that work, teams understand the architecture and decisions behind what they’ve built, along with the methods used to train and operate it, so they can run and improve their AI independently.

    Sovereignty does not always require self-hosting; sometimes it just requires deployments that meet your control requirements. For organizations that prefer a managed service, Mistral Large 4 is available through Microsoft Foundry, with private deployment options for Azure developers.

    Start building with Le Chonk

    Mistral Large 4 is available to try, deploy, and customize today.

    Get started in Mistral Studio, through the Mistral API, or on OpenRouter. For full control, download the open weights from Hugging Face and deploy on your own infrastructure with vLLM or SGLang. The model card and license have everything you need to get started — and for those who want to go deeper, a full technical report is coming soon.

    You can also try the new model in Vibe Code; give it a feature to build or a bug to fix in your own repository.

    Mistral Large 4, is also available through NVIDIA-hosted endpoints, as an optimized NVIDIA NIM container for serving with vLLM. Developers can use the open source NeMo AutoModel library for customization or the open source TensorRT LLM library for high-throughput serving.

    Enterprises can also tailor Mistral Large 4 to their proprietary data, use cases, and guardrails with Mistral Forge. Talk to our team to learn more.

    The model is available to use under the Mistral Ecosystem Open-Weight license, a modified MIT license. The model is free to use, modify, fine-tune, and commercialize, with a single condition: businesses that operate a Model-as-a-Service offering must obtain a separate commercial license from Mistral before using the model commercially.

    Starting today, Mistral Large 4 is priced at $0.70 input / M, $0.07 cached input / M, and $2.10 output / M tokens.

    We encourage the community to put Mistral Large 4 to the test. Run your own evaluations, push its limits, and share your findings with us on X, Discord, or GitHub.

    What's Next

    Mistral Large 4 will serve as the foundation for a new generation of specialized and optimized Mistral models. Building on recent releases such as Voxtral and Timestral, we'll continue expanding the frontier of open-weight AI, continuing to give developers and enterprises greater freedom to build, customize, and own their intelligence.

    Mistral Large 4

    NEW

    Open-weight hybrid instruct-and-reasoning MoE with multimodal input; unifies instruction, reasoning, and agentic capabilities in a single model, state-of-the-art among open weights on cybersecurity, finance, and manufacturing, natively fluent in 160+ languages.

    MULTIMODAL
    REASONING
    CODING
    AGENTIC
    CYBERSECURITY

    Input (/M tokens): $0.7

    Output (/M tokens): $2.1

    Original source
  • Oct 28, 2026
    • Date parsed from source:
      Oct 28, 2026
    • First seen by Releasebot:
      Nov 17, 2026
    • Modified by Releasebot:
      Nov 17, 2026
    Mistral logo

    Mistral

    October 28

    Mistral updates Large 4 reasoning controls, supporting low, medium, and high effort while rejecting none and defaulting to high.

    Mistral Large 4 (mistral-large-4, mistral-large-4-0) now supports reasoning_effort values low, medium, and high, and no longer accepts none. Requests that set reasoning_effort="none" return HTTP 400.

    Mistral Large 4 always reasons: use low for the least reasoning. If you omit reasoning_effort, the model uses high. For details, see Mistral Large 4 reasoning effort.

    API UPDATED

    Original source
  • Oct 28, 2026
    • Date parsed from source:
      Oct 28, 2026
    • First seen by Releasebot:
      Oct 29, 2026
    Mistral logo

    Mistral Common by Mistral

    v1.13.0:Add response_fromat support to model settings.

    Mistral Common releases updates with response-format support, configurable chat template defaults and reasoning effort levels, plus tokenizer and Hugging Face Hub offline fixes and a new grammar entry point.

    What's Changed

    • Honor cache_dir when listing locally cached tokenizer files by @Zakariakhchiche in #349
    • Reject multiple tool results in tokenizer v2 by @juliendenize in #339
    • Fix stale comment and positional call in v2 parallel-results tests by @juliendenize in #372
    • Add Python 3.15 support by @juliendenize in #373
    • Fix offline fallback in download_tokenizer_from_hf_hub for huggingface_hub>=1.0 (httpx) by @arthurdjn in #377
    • Fix HTML-escaped tool-call JSON in generated chat templates by @juliendenize in #375
    • Fix HF Hub offline fallback for huggingface_hub 2.x (httpx2) errors by @juliendenize in #387
    • Treat a null strict as the default in Function.from_openai by @Zakariakhchiche in #396
    • Remove legacy integer backward compatibility for SpecialTokenPolicy by @juliendenize in #402
    • Add response_format support to model settings (no version bump) by @juliendenize in #355
    • Add response format support to generated chat templates by @juliendenize in #356
    • Switch docs deploy from actions-gh-pages to official GitHub Pages Actions by @juliendenize in #403
    • Fix docs deploy: pin upload-artifact SHA instead of composite action by @juliendenize in #404
    • Migrate core runtime tests to source-relative unit modules by @juliendenize in #338
    • Drop Python 3.10 support by @juliendenize in #407
    • Raise project errors for public audio requests by @juliendenize in #342
    • Centralize base64 data URL prefix handling in mistral_common.utils.data_url by @juliendenize in #405
    • Drop the strict flag; always reject malformed data URLs (#405) by @juliendenize in #411
    • Move v1-v3 instruct tokenizer tests into the unit suite by @juliendenize in #374
    • Move v7 instruct tokenizer tests into the unit suite by @juliendenize in #388
    • Move v11 instruct tokenizer tests into the unit suite by @juliendenize in #416
    • Move v13 instruct tokenizer tests into the unit suite by @juliendenize in #417
    • Add reasoning effort levels and configurable chat template defaults by @juliendenize in #421
    • Fix typos by @DimitriPapadopoulos in #383
    • Add GrammarFactory.get_grammar_for_chat_request single-pass grammar entry point by @juliendenize in #414
    • Bump version to 1.13.0 by @juliendenize in #422

    New Contributors

    • @Zakariakhchiche made their first contribution in #349
    • @arthurdjn made their first contribution in #377
    • @DimitriPapadopoulos made their first contribution in #383

    Full Changelog: v1.12.0...v1.13.0

    Original source
  • Similar to Mistral with recent updates:

  • Oct 21, 2026
    • Date parsed from source:
      Oct 21, 2026
    • First seen by Releasebot:
      Oct 22, 2026
    Mistral logo

    Mistral

    October 21

    Mistral notes Z.ai’s GLM 5.3 Flash is now available, a multimodal model built for developer agents, visual coding, and complex tool use.

    Z.ai GLM 5.3 Flash (zai-glm-5-3-flash) is now available.

    MODEL RELEASED

    • 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.
    Original source
  • October 2026
    • No date parsed from source.
    • First seen by Releasebot:
      Oct 21, 2026
    Mistral logo

    Mistral

    Voxtral 3

    Mistral releases Voxtral 3, a voice-agent SDK and speech models for real-time multilingual conversations, combining transcription, expressive speech, reasoning and tool use. Available through hosted APIs or self-deployment, it supports over 50 languages and gives organizations control over audio data.

    Summary

    Voxtral 3 introduces a cutting-edge voice agent SDK and state-of-the-art speech models designed for real-time, multi-step conversations across 50+ languages. It combines accurate transcription, natural speech synthesis with emotional adaptability, and deep tool integration for seamless workflow automation. Available via Mistral’s EU-hosted API or self-deployment, it ensures data sovereignty while delivering low-latency, human-like interactions for contact centers, virtual assistants, and hands-free copilots.

    Speech transcription and voice generation have come a long way, but building end-to-end voice agents that can listen, reason, interact, and handle complex workflows remains difficult.

    A reliable voice agent needs to do four things very well (and simultaneously):

    • Listen accurately across languages, accents, speakers, and interruptions,
    • Reason deeply about what users mean in specific contexts,
    • Speak naturally with emotion and expression, and
    • Take action over multiple steps through connected systems.

    Getting all four to work together reliably in real time remains a challenge.

    To address this, we introduce Voxtral 3: a new family of state-of-the-art speech-to-text and text-to-speech models, alongside a voice agent SDK that brings listening, thinking, tool calling, and speaking into a single real-time streaming pipeline.

    Voxtral 3 is built to create natural conversational experiences, from contact-center agents and service desks to virtual assistants and hands-free copilots. With support for dozens of languages, expressive speech, complex reasoning, and deep integration with your own tools and data, agents built with Voxtral can interpret nuanced intent, reason through sophisticated requests, and execute multi-step actions across connected systems.

    As with other Mistral models, the Voxtral 3 family is designed for sovereign deployment. Available today through Mistral's EU-hosted API or for self-deployment on your own infrastructure, it gives organizations control over where audio is processed, how data is handled, and whether conversations are retained.

    Highlights

    • Accurate transcription across 50+ languages. The speech-to-text model, Voxtral Small Transcribe 3, is state-of-the-art in multilingual transcription, identifying who said what (diarization), using keyphrases to improve its accuracy (context-biasing). It detects when someone has finished speaking so agents can respond without delay (end-of-turn detection).
    • Natural speech that adapts to the conversation. The text-to-speech model, Voxtral Mini TTS 3, introduces new voices that adjust their emotion, pace, and emphasis based on conversational context, with support for 35+ languages and 50+ dialects.
    • Human-like intelligence at your (spoken) command. The new voice agent SDK combines listening, reasoning, tool calling, and speaking into a real-time experience. Agents can retrieve information, interact with business systems, and complete tasks during live conversations over phone or web.
    • Pro Voice for distinctly accurate voice adaptation. For professional use cases where basic voice cloning does not suffice, Voxtral’s Pro Voice adaptation is able to reproduce a speaker’s distinctive character, emotional range, and speaking style across languages, using just 30 minutes of audio.
    • Your voice AI, on your infrastructure. Use Mistral's EU-hosted API or deploy Voxtral 3 on your own infrastructure. Keep sensitive conversations within your environment and control whether any audio is retained.

    It hears 50+ languages, with state-of-the-art accuracy.

    Voxtral Transcribe 3 represents a significant milestone across transcription, diarization (who spoke when) and context biasing capabilities. The supported list of languages is expanded to 50, including 24 official EU languages and many low-resource Asian and Indic languages.

    The Small version achieves state-of-the-art results across diarization and multilingual transcription, including in English, French and Spanish. It outperforms Whisper, Scribe v2 and Gemini 3.5 Transcribe. The Mini variant is 4B parameters, enabling fast and accurate transcriptions at a price of $0.0035/min.

    Both models support context biasing, where key terms and free-form text can be given to bias the transcription and improve recall on technical terms. This enables them to be used in production settings with domain-specific vocabulary without the need for fine-tuning.

    Voxtral Realtime Transcribe 3 is a realtime transcription model built for agentic use-cases. It achieves high-quality transcriptions at sub-500ms latency in over 35 languages, including with diarization and context biasing.

    End-of-turn prediction is natively supported, incorporating both acoustic and semantic information to accurately predict when a user is finished speaking. This allows for low-latency voice agent applications, while minimizing false cut-off rates when the user is prematurely interrupted.

    It answers with the right emotion, in your or your customer’s brand's own voice and local dialect.

    Voxtral Mini TTS 3.0, available in batch and realtime, achieves state-of-the-art expressiveness and instruction following for 35+ languages and 50+ dialects. At $13 per 1M characters, Voxtral Speech 3 costs less to run than comparable models from OpenAI TTS-1, Deepgram, ElevenLabs Flash and Gemini Flash.

    The speech model implicitly interprets the semantic meaning of context better, producing natural speech across a diverse range of emotions and languages. New and improved features include voice cloning from 30+ minutes of sampled audio, custom prompt instruction following (IF), contextual TTS, and a vocabulary library, all of which serve the goal of speech reliability in deployment. Variables for Explicit IF include emotion, pacing, and emphasis.

    All natural-sounding speech relies on emotion, which determines its prosody, rhythm, pacing, emphasis, and transitions. Voxtral TTS natively interprets a prompt’s text context to determine the emotion and voice style, making the model inherently better at natural speech without requiring additional instructions or voice prompts.

    Illustrating Voxtral hill-climbing in expressiveness, numbers and unique names.

    Speech sounds local in 35 languages and over 50 dialects.

    Voxtral TTS supports over 35 languages and regional dialects. Our multilingual evals adhere to dialect differences in pronunciation, formal vs. informal register, spelling, verbosity, and locale conventions (time, dates, temperatures, addresses, currencies).

    Voxtral 3 multilingual library

    Voice Presets are available in 18 languages: Arabic (Emirati, Modern Standard), Brazilian (Portuguese), Catalan, Chinese (Mandarin), Dutch, English (American, British), French (France), German, Greek, Hindi, Italian, Korean, Norwegian, Russian, Spanish (Spain), and Vietnamese.

    The voice agent is in control of the whole conversation.

    Voxtral 3 introduces a voice agent SDK, built on LiveKit, that runs speech in, reasoning, and speech out as one streaming pipeline. Voxtral Mini Transcribe Realtime 3 listens and detects when the caller has finished speaking, a general model—like Mistral Small—reasons and calls tools, and Voxtral Speech 3 replies in the tone of the conversation so far. No stage waits for the one before it, so each turn completes in under one second, and the agent stops talking when the caller cuts in.

    During a call, the agent can verify a caller’s identity, check a bill, or book an appointment, using function calling, web search, and retrieval over your own systems. It answers and places calls over SIP, runs in browsers and apps over WebRTC, and connects to the telephony, IVR, and contact-center software you already use.

    Each stage is a separate model, so you can change the LLM, the voice, or the language without rewriting the agent, and run the whole stack on Mistral's API in Europe or on your own infrastructure.

    Get started.

    Voxtral 3 is available today through the Mistral API, hosted through our European, US, and global endpoints.

    You can test out voice agents, audio transcription, speech generation, and voice cloning in the Studio playgrounds.

    Model Price Voxtral Mini Transcribe 3 (batch) $0.0035/min Voxtral Mini Transcribe Realtime 3 $0.008/min Voxtral Small Transcribe 3 $0.0040/min Voxtral Speech 3 (batch and realtime, same API) $13/1M chars Voice agent API with Mistral Small 4 for reasoning 0.04$/min Voice agent API with any other reasoning model 0.08$/min

    To run the full stack on-premises or in a private cloud, talk to our team.

    Voxtral Small Transcribe 3

    NEW

    Highest accuracy speech transcription model. State-of-the-art multilingual transcription and diarization, with support for over 50 languages and context biasing.

    TRANSCRIPTION

    Audio

    $0.004

    /min

    Voxtral Mini Transcribe 3

    NEW

    Efficient, low-latency speech transcription model. Supports multilingual transcription, diarization and context biasing in over 50 languages.

    TRANSCRIPTION

    Audio (Batch)

    $0.0035

    /min

    Voxtral Mini Transcribe Realtime 3

    A realtime transcription model built for agentic use-cases, delivering high-quality transcriptions at sub-500ms latency across 40+ languages.

    TRANSCRIPTION

    Audio

    $0.008

    /min

    Voxtral Speech 3

    State-of-the-art expressiveness and instruction following across 35+ languages and 50+ dialects

    VOICE

    Input

    $13

    /1M chars

    Original source
  • Oct 19, 2026
    • Date parsed from source:
      Oct 19, 2026
    • First seen by Releasebot:
      Oct 22, 2026
    Mistral logo

    Mistral

    October 19

    Mistral releases the Voxtral 3 audio model family in public preview, bringing new speech-to-text and text-to-speech capabilities, including voice cloning and real-time transcription. Mistral also launches managed Voice Agents for real-time voice pipelines across telephony, browser, and custom-audio interfaces.

    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 and the audio models pricing.

    MODEL RELEASED

    • Voxtral TTS 3 (voxtral-mini-tts-3-0): text-to-speech with zero-shot voice cloning, streaming, and contextual delivery. See the model card.
    • Voxtral Mini Transcribe 3 (voxtral-mini-transcribe-3-0): offline transcription and diarization with timestamps. See the model card.
    • Voxtral Mini Transcribe Realtime 3 (voxtral-mini-transcribe-realtime-3-0): real-time transcription. See the model card.
    • Voxtral Small Transcribe 3 (voxtral-small-transcribe-3-0): offline transcription and diarization with timestamps. See the model card.

    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 and Voice Agents pricing.

    OTHER

    Original source
  • Oct 15, 2026
    • Date parsed from source:
      Oct 15, 2026
    • First seen by Releasebot:
      Oct 16, 2026
    Mistral logo

    Mistral

    Introducing Timestral

    Mistral releases Timestral, its first family of time-series foundation models, with four sizes for probabilistic, multivariate forecasting and support for past-only and known-future covariates. The models lead public benchmarks, run on standard CPUs, and are available to try through a published GitHub repository.

    Today, we’re releasing Timestral, our first family of time-series foundation models. Our largest model, at just 209M parameters, ranks first on all three of the field's main public benchmarks, GIFT-Eval, TIME, and fev-bench, outperforming the previous state of the art with only two-thirds of the model parameters.

    Timestral delivers probabilistic, multivariate forecasts and supports past-only and known-future covariates. Our family of models sits on the Pareto front of forecasting accuracy versus model size, delivering leading performance at each size. Results have been submitted to the GIFT-Eval and TIME leaderboards.

    Time Series Foundation Models (TSFM)

    Time series is one of the most common data types in enterprise, from sensor telemetry to observability metrics. Forecasts built on it drive decisions in capacity planning, energy, supply chains, risk and maintenance.

    Industrial plants forecast process sensors against the production schedule, catch equipment drift before it fails, and plan capacity and spare parts across thousands of assets. Finance teams forecast flows, volumes, and volatility to inform positions, hedging, and treasury across thousands of accounts.

    Classical models such as ARIMA and linear regression remain well-established tools for forecasting time series. These models are usually fitted to a single series and learn only from that series' history.

    Deep learning models such as PatchTST add the capacity to learn nonlinear patterns across many related series, but they are still typically trained from scratch on each dataset, so what they learn doesn't carry over to new domains.

    Time series foundation models bring the pre-training approach to time series. Trained on a large and diverse corpus, our model learns patterns that transfer across series, forecasting zero-shot on data it has never seen, even in domains where data is scarce.

    Benchmark Results

    We evaluated Timestral on the three main public forecasting benchmarks: GIFT-Eval, TIME and fev-bench. Among pre-trained, zero-shot models, it ranks first across all five metrics. These results come from a single checkpoint, without ensembling.

    GIFT-Eval, top ten of 83 comparable pretrained and zero-shot entries. Lower is better.

    TIME, top ten of the 35 leaderboard entries. Lower is better on both metrics.

    fev-bench, top ten of the 37 leaderboard entries, by SQL skill score. Higher is better.

    Family of models optimised for every specification

    Timestral has been optimised as a family of four models: 5M, 28M, 123M and 209M parameters. The larger models improve forecast performance, while the smaller ones are faster and cheaper. All four share the same architecture and training recipe. Every size sits on the Pareto front of accuracy versus model size: for each model size, no public model forecasts better.

    Score vs model size on GIFT-Eval, TIME and fev-bench (shown as 1 − skill score). Lower is better. Our 28M model beats Chronos-2, about four times larger, on GIFT-Eval and TIME. Our 209M model beats TimesFM-3 and Falcon-2.0 with fewer parameters.

    Timestral does not need a GPU at inference. On a standard server CPU with 8 threads, forecasting one day ahead from three weeks of 15-minute history takes 10 ms with the 5M model and 140 ms with the 209M model. Smaller models trade some accuracy for speed: the 28M model is five times faster than the 209M model, and the 123M model comes within 0.2% of the 209M score at three quarters of its latency.

    An industrial example

    Consider an industrial plant with thousands of sensors across its equipment. Timestral can improve the forecasts on that data in several ways.

    Multivariate forecasting

    A pump's vibration, temperature and pressure don't move independently. A rise in one often explains or causes a change in another. Timestral models related series jointly, so it can use those dependencies instead of forecasting each sensor in isolation.

    Probabilistic forecasting

    A single predicted value may not show the full picture. Timestral predicts a full set of quantiles, so teams can see how wide the uncertainty is, set alert thresholds that reflect it, and plan maintenance around risk rather than a single number.

    Categorical covariates

    Some covariates take values from a set of categories rather than continuous numbers, such as the day of the week, or a machine's state (running, idle, under maintenance). Encoding these as numbers would impose an order or scale they don't have. Timestral takes them as categorical inputs directly to improve forecasts.

    Known-future covariates

    Some drivers of sensor behaviour are known in advance, such as production schedules, planned shutdowns, weather forecasts. Timestral can condition on them to improve forecast quality, so a scheduled load increase is expected rather than flagged as a surprise.

    From Model to System

    Timestral is a numerical model that can run inside existing monitoring pipelines, or be paired with our language models in multi-modal agentic workflows.

    In the plant, that could look like this:

    1. Timestral running continuously on the plant's sensor streams, detecting anomalies as they occur.
    2. When it flags an anomaly on a pump, an agent embeds the signals and compares against embeddings of past sensor data to find similar historical failures.
    3. The agent then starts sub-agents to run a root cause analysis, drawing on context from that plant, such as maintenance records, shift logs and operator notes, equipment manuals, past work orders, and readings from related equipment.
    4. The agent drafts a suggested diagnosis based on the findings, with supporting evidence and possible next steps.

    The plant manager gets a diagnosis and a plan backed by the sensor data, rather than just an alert.

    Key technical contributions

    Several ideas have been combined to produce these results. Each will be described in detail in the technical report.

    1. Decoupled Horizon Forecasting. Causal decoders trained with contiguous patch masking, as in Toto 2.0 and TimesFM-3, mask out the values to be forecast, so they look like missing values and are also lost as context for later forecasts. Our approach, DHF, separates the two: the full history stays in the sequence, and learned forecast tokens read the history before them. Every observed value thus serves as both target and context, true gaps stay flagged as missing, and training horizon lengths and missing-data augmentation can be tuned independently of each other.
    2. Covariate-aware attention. Each input series is assigned a role: target variate, past-only covariate or known-future covariate. Known-future covariates are visible across the whole window while targets stay causal, and information flows only from covariates to targets. Timestral can therefore use a holiday calendar or a weather forecast without leaking a target's own future into its forecast.
    3. Synthetic-first data. The majority of the training data was synthetic. We first generated univariate processes with coherent temporal dynamics, including trends, seasonality and diverse noise patterns. We then used a multivariate synthesizer to construct multivariate series by introducing nonlinear dependencies between these processes.
    4. Two-stage training. Timestral is first pre-trained at scale on purely synthetic data with context size 8192, then post-trained on a mix of real and synthetic data with a 4x longer context. The objective is a pinball loss over 21 quantiles, a discrete approximation of CRPS.

    A model that can be adapted

    Public benchmarks measure general performance, but production data can be significantly different in behavior and structure. It may also shift over time as markets change regime, equipment degrades and demand patterns move. The lasting value lies in being able to adapt a model to in-domain data and keep retraining it as the distribution moves.

    For organisations that want to go further, Timestral is available through Mistral Forge, the same training pipeline we use for our language models. It supports:

    • Zero-shot inference out of the box.
    • Architectural extensions, such as domain-specific covariates.
    • Fine-tuning on proprietary data, specialising to a domain while preserving general performance.
    • Control over the training data, including adjusting the training mix, replaying pre-training data, and generating synthetic data samples to reflect the target domain. Full control over the data ensures that nothing we backtest was seen during pre-training.

    The full pipeline can be deployed on our clients’ hardware, including air-gapped environments.

    What’s next

    We have published a GitHub repository so you can run Timestral on your own data. A technical report with the full results will follow. For more specialised forecasting problems, or to collaborate on research, please reach out to us.

    Timestral

    • Type: Time-series foundation model (TSFM)
    • Parameters: 5M, 28M, 123M and 209M
    • Context length: 32,768 time steps
    • Output: Probabilistic forecasts, 21 quantiles
    • Inputs: Multivariate targets, past-only and known-future covariates
    • Deployment: On-prem, VPC or air-gapped
    • Licence: Modified MIT
    Original source
  • Oct 6, 2026
    • Date parsed from source:
      Oct 6, 2026
    • First seen by Releasebot:
      Oct 6, 2026
    Mistral logo

    Mistral

    Mistral Large 4

    Mistral launches a public preview of Mistral Large 4, its largest and most capable open-weight model yet. Available now on Mistral Studio, the multimodal model brings major gains in coding, agentic workflows, cybersecurity, finance, law, and visual understanding, with weights coming later this month.

    Today, we’re launching a public preview of Mistral Large 4

    Unofficially ML4, very officially: le Chonk. ML4 pushes the frontier of open-weight performance. You can try the preview API today on Mistral Studio. Weights drop end of this month.

    Frontier performance

    ML4 is a 1 trillion-parameter natively multimodal model with 49 billion active parameters. It is our largest and most capable model to date, and it continues to improve rapidly as we refine it.

    The model demonstrates exceptional performance across coding, agentic workflows, and multimodal understanding. It already achieves performance competitive with the strongest open-source models globally, while significantly outperforming any open-weight model developed in the US or Europe. On critical enterprise workloads, including cybersecurity, finance and law, we find it to be state-of-the-art among open models. In some domains such as visual grounding, it goes further still, surpassing even frontier closed models.

    We will release the weights by the end of the month. Until then, we are red-teaming the model in real-world settings with cybersecurity leaders, vetted partners, and state authorities, who will access the same model with reduced moderation and expanded cyber capabilities.

    Capabilities include:

    • Coding - DeepSWE
    • Coding - Terminal Bench 4.0
    • Cyber
    • Agentic behaviour
    • Finance Agent
    • Harvey's Legal Agent
    • Grounding

    Forged in Europe. Built for AI sovereignty.

    ML4 was trained from scratch on 3,800 NVIDIA Grace Blackwell GPUs in Mistral’s own datacenters in Europe. The public preview is served on that same infrastructure. It is a significant milestone in our long-term investment across infrastructure, research, and product development: state-of-the-art performance in critical verticals, delivered through open weights, designed to give customers control over their AI.

    This is particularly important in cybersecurity, where provider-level refusals can block legitimate vulnerability research and incident response, and where losing access to a capability mid-incident can itself become a critical security risk. ML4 pairs top-tier cyber performance with open weights and self-deployment, giving organizations both the capability and the autonomy to run advanced security work under their own policies.

    The model will be available across multiple regions worldwide, including a European deployment that Mistral operates end-to-end, independently of other digital service providers and under European law. Fun fact: a significant share of ML4’s training data was multilingual, spanning more than 160 languages, including every official language of the European Union.

    We’ve been working closely with leading enterprises across the world in finance, engineering, manufacturing, logistics, pharmaceuticals, science, shipping, public sector, and other mission-critical industries to train ML4. In fact, the model uses the same training, customization, and RL environment we offer our customers through Mistral Forge.

    Try it today

    There is still more to come. As we work toward releasing the weights, we will share further details on the model architecture, additional benchmarks, and our post-training methodology.

    This model will also serve as the foundation for a new generation of specialized and optimized Mistral models. In the meantime, we invite you to try the preview API and share your feedback with us on social media.

    Capabilities deep-dive

    Cybersecurity

    ML4 is one of the world's strongest AI models for cybersecurity. On the Artificial Analysis Cyber Index, an independent evaluation of how well AI models find and fix security flaws in real software, it ranks among the top five models globally and leads open-weight models developed outside China by a wide margin. On one of the index's tests, which asks a model to reproduce a real vulnerability in open-source software and then patch it, ML4 scores 82%, the highest of any model. It also solves 93% of the challenges in Cybench, a set of 40 exercises drawn from security competitions, one of the highest scores reported for an open-weight model.

    That top score reflects a practical advantage. Several leading closed models, including Claude Opus 5.5 and GPT-6 Astra, score near zero on the same test because they refuse to perform the task. Yet defending software often starts with proving that a flaw is real, exactly the kind of work safety filters in closed models can block. This matters even more as threat actors increasingly jailbreak those same models to support offensive cyber activity: defenders need systems that can match those capabilities without being constrained by the same refusals. ML4 can do that work, and its capabilities extend beyond what it was explicitly trained for: in internal testing, it proved useful for analysing malware, prioritising vulnerabilities, and writing detection rules. For organisations that need sovereign, auditable AI for security operations, it will be able to run on private cloud or on-premise.

    Agentic coding

    ML4 excels across software engineering, repository understanding, and complex terminal workflows, scoring 61.7% on DeepSWE v1.1, 59.4% on SWE-Atlas-QnA, and 28.3% on Terminal-Bench 4. Its combined Coding Agent Index score of 49.8% places it ahead of DeepSeek V4 Pro 0813 and Qwen3.8 Max.

    We also ran a blind human evaluation with Surge AI on coding quality: professional annotators rated model outputs on a 1–5 scale, with model identities hidden. ML4 Preview ranked second of five models (3.74), ahead of Kimi K3 (3.59), GLM-5.3 (3.60) and GLM-5.2 (3.40), and behind only Claude Opus 5 (4.22).

    Agentic Workflows

    ML4 runs general-purpose agents that gather information, use tools, and produce finished deliverables across complex workflows. On AutomationBench — 657 business workflows across apps like Gmail, Google Sheets, Slack, and Salesforce — it scores 59.9%, ahead of Kimi K3, MiMo-V2.6-Pro, and DeepSeek V4 Pro.

    It's just as strong on the professional deliverables that knowledge work actually produces: spreadsheets, slides, and PDFs. On AA-Briefcase, which evaluates long-horizon knowledge work, it reaches 1,393 Elo, ahead of DeepSeek V4 Pro.

    Multimodal

    ML4 is a step change in the ability of our models to understand images. It reasons powerfully across complex documents, charts, and natural images, and brings vision to the industries where perception is critical such as engineering, manufacturing, and earth observation.

    The model can further combine visual grounding with agentic capabilities: from inspecting gigapixel satellite imagery — helping disaster-response teams act when time counts — to analyzing engineering-drawings — zooming in, inspecting, and verifying until the answer is exact. In our demos above, ML4 grounds dense natural scenes, verifies mechanical parts in technical drawings, retrieves evidence from PDFs, and scans massive geospatial images for the hardest-to-find objects.

    On visual grounding particularly, we find ML4 to be one of the most capable models we tested, for instance surpassing GPT-6-Astra on Dense 200 (42% vs 41%).

    Science and Math

    ML4 brings strong scientific capabilities, built by combining AI-driven methods with our researchers' expertise in mathematics, physics, and chemistry.

    It's highly proficient at agentic coding for scientific tasks such as data analysis, modeling, and simulating physical reality, which lets researchers focus on the questions rather than the plumbing. In benchmarks, ML4 is state of the art on SciCode-Verified among open-weight models. In practice, it can generate a full Hartree–Fock simulation in one shot — a complex, multi-step chemistry task built from a series of advanced routines.

    ML4's math is stronger too, in both formal reasoning and applied mathematics. In our human evaluations it reasons more precisely and more with more structure than GLM-5.3, and it can sustain long, domain-specific applied-mathematics tasks, including work relevant to frontier theoretical physics.

    Together, these capabilities make ML4 a strong research assistant across the full technical workflow — from the first question to the final result.

    Knowledge Work

    ML4 is our most capable model for the real-world tasks which professionals handle every day. It can create, edit and fix complex spreadsheets and documents, showing exemplary performance on both legal and financial benchmarks.

    Notably, we evaluated ML4 through third party evaluators (vals.ai) on representative tasks for both legal and financial tasks, finding the model exceeds GPT-6-Astra in both cases. On HarveyAI’s Legal Agent benchmark, ML4 outperforms all open-source models.

    Financial analysis demands precision and the ability to synthesize information from multiple sources, a process that remains time-consuming at many financial institutions today. In this demo, ML4 compared to other top OSS models take on the same multistep corporate finance challenge, searching through public company filings and financial reports, such as those available via EDGAR and equivalent European databases. An animated semantic map traces each model's journey toward a solution, highlighting every document retrieved along the way. Each track's position reflects the evidence gathered, the results of calculations, and the questions that remain unresolved. Viewers can follow how the investigations unfold and compare the distinct paths each model takes before arriving at its final answer.

    Model Safety

    ML4 has saturated our benchmarks on robustness to indirect prompt injections, putting it at the frontier of OSS models (compared to GLM-5.2, GLM-5.3, Kimi-K2.6, Kimi-K3, DS-V4-Pro-0813). On Lakera’s public B3 AI Security Benchmark, ML4 resists 93.3% of attacks – we see no higher scores among competitors.

    ML4 also engages more responsibly with users than any of our previous models. We highlight our results on the KORA Benchmark, where ML4 again sits at our highest measured score among OSS models (1.691, with 2 being the maximum denoted as “Exemplary”).

    Of particular relevance is the model’s propensity to refuse malicious requests regarding cybersecurity. Despite strong performance on Cyber benchmarks, the average refusal rate of the model on cyber prompts from JailbreakBench, StrongREJECT, and AgentHarm is higher than all OSS models.

    Human Evaluation

    We ran an internal evaluation in which expert annotators across coding, computer-aided design (CAD), finance, mathematics and physics compared Mistral Large 4 with GLM-5.3. ML4 was preferred in CAD and STEM, while performing on par or close to GLM-5.3 in finance and coding.

    Reinforcement learning at scale

    Base models are improving fast, and our post-training has to keep pace. A recipe tuned for yesterday's model leaves capability on the table with today's frontier, because ground truth samples that once pushed a model to its limits won’t anymore. We use Reinforcement Learning (RL) because it adapts as the model does: we train on the outcomes of the model's own attempts, and we can raise the difficulty and the breadth of the tasks as it gets stronger.

    Our RL library was designed to make new environments easy to add and train at scale. A shared, composable interface allows a single training run to combine tasks ranging from single-turn chat and complex scientific problem solving to safety alignment, factuality, and long-horizon tool use. These environments share scaffolds and resources such as code sandboxes, web search, and external APIs. The same composability extends to verification, with reward models, unit tests, LLM judges, and static checks combined as needed for each task.

    At runtime, an autoscaling fleet of actors generates tens of thousands of rollouts in parallel while model training proceeds asynchronously. The generation and training pipeline is optimized for long trajectories, supporting rollout budgets of millions of tokens across multiple compactions while keeping staleness low. Novel methods and optimizations across both stages minimize off-policy drift and enable stable RL over long horizons.

    At our current scale (3k GPUs), a single training run produces roughly 33 billion tokens per day, of which around 16 billion are trainable completion tokens after filtering and masking. We can see the run progress directly in the training rollouts: training rewards rise across several representative environments as the policy learns to solve increasingly complex tasks.

    The improvements are not specific to the environments we train on; they transfer to downstream evals, and the final model owes them to both post-training stages (supervised fine-tuning and RL), as shown in the charts.

    What comes next

    This is only the beginning. ML4 is the first milestone on the roadmap funded by our €3 billion Series D — the largest equity round ever raised by a European technology company. That capital is already being put to work: we are significantly scaling up our compute capacity in our own European datacenters, and much more is coming online in the months ahead.

    More compute means more training. The reinforcement learning run behind this preview is still in flight, and the model is showing no signs of saturation — there is substantial headroom ahead. As we scale up training on our expanded infrastructure, we expect large and rapid improvements in the weeks and months to come.

    We will release the weights by the end of the month, along with more details on the architecture, additional benchmarks, and our post-training methodology. And ML4 is only the foundation: it will serve as the base for a new generation of specialized and optimized Mistral models, built for the industries and workloads our customers care about most.

    The pace of progress from here will be fast. Stay tuned.

    Original source
  • Oct 5, 2026
    • Date parsed from source:
      Oct 5, 2026
    • First seen by Releasebot:
      Oct 6, 2026
    Mistral logo

    Mistral

    October 5

    Mistral launches Mistral Large 4 in Public Preview with open-weight multimodal support and a 1M context window.

    Mistral Large 4 (mistral-large-4) is now available in Public Preview. For details, see the Mistral Large 4 model card.

    MODEL RELEASED

    • 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.
    Original source
  • Sep 28, 2026
    • Date parsed from source:
      Sep 28, 2026
    • First seen by Releasebot:
      Sep 30, 2026
    Mistral logo

    Mistral

    September 28

    Mistral deprecates OCR 4.0 and Leanstral 1.5 and points users to OCR 4.1 and GLM 5.3 at the same price.

    DEPRECATED

    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

    Original source
  • Sep 27, 2026
    • Date parsed from source:
      Sep 27, 2026
    • First seen by Releasebot:
      Sep 30, 2026
    Mistral logo

    Mistral

    September 27

    Mistral notes Z.ai GLM 5.3 is now generally available.

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

    MODEL RELEASED

    Original source
  • Sep 22, 2026
    • Date parsed from source:
      Sep 22, 2026
    • First seen by Releasebot:
      Sep 22, 2026
    Mistral logo

    Mistral Common by Mistral

    v1.12.0: Agnostic validation mode

    Mistral Common releases 1.12.0 with broader validation, safer multimodal handling, and chat template improvements. It also fixes image and audio edge cases, tightens tool call and message checks, and refreshes docs, contribution guidance, and release tooling.

    What's Changed

    • Add comment and coverage guidelines to AGENTS.md by @juliendenize in #278
    • Allow overriding auto-detected plain thinking support in chat template generation by @juliendenize in #285
    • Rename LICENCE to LICENSE by @juliendenize in #286
    • Set reference audio filename for speech multipart uploads by @Mr-Neutr0n in #284
    • Keep at least one pixel per axis when resizing an image by @arthi-arumugam-git in #289
    • Reject AudioConfig where audio_length_per_tok floors to zero by @shoemoney in #292
    • Deprecate Tokenized.text in favor of tokenizer.decode by @juliendenize in #282
    • fix(tokenizer): validate ImageConfig fields to prevent ZeroDivisionError by @shoemoney in #294
    • Update uv tooling to 0.12.6 by @juliendenize in #297
    • Parallelize unit and integration test suites by @juliendenize in #298
    • Reject null tool call IDs before encoding by @juliendenize in #293
    • Deprecate direct continue_final_message on ChatCompletionRequest by @juliendenize in #303
    • Improve typing at dynamic tokenizer boundaries by @juliendenize in #296
    • Align tokenizer type contracts by @juliendenize in #299
    • docs: fix typo in experimental usage guide by @ManoharPaturi in #306
    • Add contribution guidelines and templates by @juliendenize in #314
    • Simplify issue creation by @juliendenize in #315
    • Switch documentation from MkDocs to Zensical by @juliendenize in #318
    • Fix docs deploy: replace non-existent zensical gh-deploy with actions-gh-pages by @juliendenize in #319
    • Fix actions-gh-pages ref to real v4 commit SHA by @juliendenize in #320
    • Improve docstrings across the codebase by @juliendenize in #317
    • Add agnostic validation mode by @juliendenize in #295
    • Raise InvalidToolCallError on malformed v2-v7 tool call outputs by @Chocolatine75 in #310
    • Raise InvalidMessageStructureException for invalid continue_final_message by @juliendenize in #325
    • fix(image): preserve float32 dtype and avoid float64 promotion in normalize (#316) by @andriiorap in #321
    • Allow null assistant content in OpenAI requests by @shubhangi013 in #327
    • Fix image URL loading crashes and add a download timeout by @BarneyChambers in #308
    • Add test-compat label workflow: vLLM + Transformers compat tests on CPU by @juliendenize in #334
    • Release 1.12.0 by @juliendenize in #326

    New Contributors

    • @Mr-Neutr0n made their first contribution in #284
    • @arthi-arumugam-git made their first contribution in #289
    • @shoemoney made their first contribution in #292
    • @ManoharPaturi made their first contribution in #306
    • @Chocolatine75 made their first contribution in #310
    • @andriiorap made their first contribution in #321
    • @shubhangi013 made their first contribution in #327
    • @BarneyChambers made their first contribution in #308

    Full Changelog: v1.11.7...v1.12.0

    Original source
  • Aug 30, 2026
    • Date parsed from source:
      Aug 30, 2026
    • First seen by Releasebot:
      Sep 1, 2026
    Mistral logo

    Mistral

    August 30

    Mistral OCR 4.1 is now generally available, expanding the company’s OCR offering.

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

    MODEL RELEASED

    Original source
Releasebot

Curated by the Releasebot team

Releasebot is an aggregator of official release notes from hundreds of software vendors and thousands of sources.

Our editorial process involves the manual review and audit of release notes procured with the help of automated systems.