# Mistral releases 1T-parameter multimodal model with open weights

Digest AI · Generative AI & Models · published 2026-10-29T00:00:00Z · updated 2026-10-29T15:25:27Z

Canonical: https://digestai.news/story/mistral-releases-1t-parameter-multimodal-model-with-open-weights

## Summary

Mistral released **Mistral Large 4**, a 1 trillion-parameter multimodal model with 49 billion active parameters and a 1 million-token context window. The model supports text and image inputs and generates text, released under the **Mistral Ecosystem Open-Weight License**, an MIT derivative with usage restrictions. It excels in cybersecurity, coding agents, and multimodal tasks, scoring **82%** on the **Artificial Analysis Cyber Index**—the highest reported score for reproducing and patching real vulnerabilities in open-source software. Closed-source models typically score near zero on this test due to refusal policies, while Mistral Large 4 handles such tasks without explicit training, according to the company’s internal benchmarks.

Mistral Large 4 combines **Mixture-of-Experts (MoE)** architecture with sparse attention mechanisms, including a **Pixtral vision encoder** for image processing. The model’s text decoder has **72 layers**, a hidden dimension of **6,144**, and **64 attention heads**, while the Pixtral encoder uses **14×14 patches** and a **patch-merge projector**. Benchmarks show strong performance in coding (**61.7%** on **DeepSWE v1.1**) and agentic workflows (**49.8%** on **Coding Agent Index**), surpassing competitors like **DeepSeek V4 Pro 0813** and **Qwen3.8 Max**. Mistral positions the model for defensive security work, noting that open weights enable custom deployment and fine-tuning without vendor lock-in. The model is available via **Hugging Face**, with **GGUF quantizations** provided by **Unsloth** and **llama.cpp**, and supports deployment on **NVIDIA Blackwell GPUs** via **vLLM** and **DSpark** for accelerated inference.

## Key points

- Mistral Large 4 is a 1T-parameter multimodal model with 49B active parameters, released under the Mistral Ecosystem Open-Weight License
- Scores 82% on Artificial Analysis Cyber Index, highest for reproducing and patching real software vulnerabilities, per Mistral’s internal benchmarks
- Supports text + image inputs, excels in coding (61.7% on DeepSWE v1.1) and agentic workflows, with GGUF quants and Hugging Face deployment options

## Why it matters

Open multimodal models like Mistral Large 4 democratize advanced security and coding tools, enabling custom fine-tuning for niche use cases without vendor restrictions. Its cybersecurity benchmarks and multimodal capabilities could accelerate research in vulnerability analysis and agentic workflows.

## Sources

1. [Welcome Le Chonk: Mistral Large 4 on Hugging Face](https://huggingface.co/blog/mistral-large4) (Hugging Face, 2026-10-29, primary source)
2. [Paris-based Mistral releases "Le Chonk," a strong coding model that still trails China overall](https://the-decoder.com/paris-based-mistral-releases-le-chonk-a-strong-coding-model-that-still-trails-china-overall) (The Decoder, 2026-10-29)
3. [Mistral’s 1-Trillion-Parameter Le Chonk Checkpoint Lands on Hugging Face](https://unite.ai/mistrals-1-trillion-parameter-le-chonk-checkpoint-lands-on-hugging-face) (Unite.AI, 2026-10-29)
4. [Mistral releases open weights for its 1 trillion-parameter Large 4 model](https://thenextweb.com/news/mistral-large-4-open-weights-release) (thenextweb.com, 2026-10-29)
5. [Mistral releases the weights for Mistral Large 4, or Le Chonk, and says it rivals top open models such as GLM-5.3 on the Artificial Analysis Coding Agent Index](https://zdnet.com/innovation/mistral-large-4-weights-release) (zdnet.com, 2026-10-29)
6. [mistralai/Mistral-Large-4-1T-A52B · Hugging Face](https://huggingface.co/mistralai/Mistral-Large-4-1T-A52B) (huggingface.co, 2026-10-29, primary source)
7. [Mistral Says Its AI Model ‘Le Chonk’ Has Leapfrogged China in the Open-Weight Race](https://wired.com/story/mistral-says-le-chonk-surpasses-chinese-models-cybersecurity) (Wired AI, 2026-10-29)
8. [Le Chonk: Built for Coders, Trusted by Defenders.](https://mistral.ai/news/mistral-large-4-oss) (Mistral AI, 2026-10-29, primary source)

Part of the developing story: [Mistral's Rise in Open AI Development](https://digestai.news/thread/mistral-and-nvidia-collaborate-to-integrate-ai-models-into-engineering-workflows) (2 stories)

## Cite

Digest AI, "Mistral releases 1T-parameter multimodal model with open weights", 29 October 2026, https://digestai.news/story/mistral-releases-1t-parameter-multimodal-model-with-open-weights

---

Written by Digest AI's editorial model from the linked sources; the sources are the record. Headlines, digests and key points are written by Digest AI and may be quoted with a link to the story page. Linked articles belong to their publishers. Terms: https://digestai.news/terms#reuse
JSON: https://digestai.news/story/mistral-releases-1t-parameter-multimodal-model-with-open-weights.json