Mistral Advanced Available in FlowHunt

Mistral Large 4

Mistral Large 4 is an Advanced-tier AI model developed by Mistral and launched as a public preview on October 6, 2026. Until the weights are released it is a closed-weight model, available through the preview API on Mistral Studio. The 524K-token context window is enough for long technical documents and sizeable codebases in a single prompt.

Read more
Mistral describes Large 4 as a 1 trillion-parameter natively multimodal model with 52 billion active parameters, and its largest and most capable model to date. It was trained from scratch on NVIDIA Grace Blackwell GPUs in Mistral’s own datacenters in Europe, with multilingual training data spanning more than 160 languages, including every official language of the European Union. Mistral announced that weights will be released at the end of October and positions the model for sovereign deployments, with a focus on coding, agentic workflows, multimodal understanding and cybersecurity. During the preview, vetted partners and state authorities test a version with reduced moderation and expanded cyber capabilities.
Try in FlowHunt Back to AI Leaderboard

Benchmarks

Each score appears once, with its rank among tracked models and who measured it.

Compared with Advanced peers on the same scale
  • Mistral Large 4this model
  • Claude Sonnet 5.5this model leads on 0 of 4
  • GPT-6 Lunathis model leads on 2 of 3
Verified independent evaluator, counted in the ranking Self-reported published by Mistral, shown but not ranked
Overall

AA Intelligence Index

Independent composite of agent, coding, knowledge and science tests by Artificial Analysis (v4).

Source ↗

Mistral Large 4 38.4 Claude Sonnet 5.5 56.0 GPT-6 Luna 38.1
#21of 45 Verified
Coding

DeepSWE

113 long-horizon software tasks written from scratch, graded by hidden tests.

Source ↗

Mistral Large 4 61.7% Claude Sonnet 5.5 71.0% GPT-6 Luna No score
Not ranked Self-reported
Agents

Terminal-Bench 4.0

Multi-step professional tasks done by an agent in a command-line terminal.

Source ↗Also: Self-reported 28.3%

Mistral Large 4 26.8% Claude Sonnet 5.5 63.6% GPT-6 Luna 12.6%
#14of 36 Verified
Reasoning

Humanity's Last Exam

Expert-written questions across 100+ subjects, no tools.

Source ↗

Mistral Large 4 35.0% Claude Sonnet 5.5 55.0% GPT-6 Luna 38.5%
#30of 44 Verified

What to Use Mistral Large 4 For

Agentic Coding

Mistral says Large 4 is strong at software engineering, repository understanding and complex terminal workflows.

Cybersecurity & Technical Work

Mistral highlights vulnerability analysis, malware analysis and writing detection rules, and says it is built for security teams that need self-deployable capability once the weights ship.

Knowledge Work in Finance & Law

It creates, edits and fixes complex spreadsheets and documents, and Mistral reports strong results on legal and financial tasks.

Multimodal & Document Understanding

As a natively multimodal model, it reasons across documents, charts and natural images, and Mistral highlights visual grounding in areas such as engineering drawings and satellite imagery.

Strengths and limitations

Strengths

  • Strong agentic coding and terminal workflows
  • Cybersecurity capability with open weights announced for self-deployment
  • Natively multimodal, including visual grounding
  • Multilingual training across more than 160 languages
  • Trained and served on European infrastructure

Limitations

  • Public preview, so the model is still being refined and Mistral says further details are to come
  • Weights are not yet released, so it cannot be self-hosted or fine-tuned until the announced end-of-October release
  • Architecture details and additional benchmarks have not been published yet

Use it in FlowHunt

Use Mistral Large 4 in any flow, next to your own tools and knowledge sources.

  1. 1Open a flow, or create one from a template
  2. 2Add an AI Agent or Generator component
  3. 3Pick Mistral Large 4 in the model selector

Related AI Models

About Mistral Paris, France · Founded 2023 · mistral.ai

Mistral AI is a French AI company founded in April 2023 by Arthur Mensch (formerly of Google DeepMind), Guillaume Lample, and Timothée Lacroix (both formerly of Meta AI Research). The company positions itself as a European champion in generative AI, combining a commercial API business with strategic open-weight releases. Mistral's flagship Large series and Apache 2.0-licensed models have made it the most prominent non-US, non-Chinese provider on most AI leaderboards. The company maintains EU headquarters in Paris, emphasises GDPR and EU regulatory compliance, and has articulated a distinct regulatory philosophy calling for lighter regulation of open-weight models compared to closed API systems.

mistral.ai ↗

Browse the Full AI Model Leaderboard

Compare every tracked model on coding, agent and reasoning benchmarks, with each score marked verified or self-reported.

← Back to AI Leaderboard Try in FlowHunt

Frequently asked questions