Mistral Large 4 open weights: Europe's 1T MoE drops this month
Mistral AI announced Mistral Large 4 on October 6, 2026 — a 1 trillion parameter natively multimodal mixture-of-experts model the company calls 'le Chonk'. The public preview API is live on Mistral Studio today. Open weights are committed for end of October 2026, with red-teaming by cybersecurity leaders and state authorities already underway. At 52B active parameters, it leads open-weight models outside China on cybersecurity, finance, and manufacturing benchmarks.
- 1T MoE architecture — Mistral Large 4 is a 1 trillion total parameter, natively multimodal mixture-of-experts model with 52 billion active parameters per token (including embeddings and output layers). Announced October 6, 2026.
- Open weights by end of October — Mistral committed to releasing weights by end of October 2026. The preview API is available now on Mistral Studio and the Mistral API (model ID: mistral-large-4).
- Cybersecurity SOTA among Western open weights — On Artificial Analysis Cyber Index, ML4 ranks top 5 globally and leads open-weight models developed outside China. Scores 82% on reproduce-and-patch vulnerability test (highest of any model) and 93% on Cybench. Closed models like Claude Opus 5.5 and GPT-6 Astra score near zero on the same test due to refusals.
This article is based on Mistral's official announcement (October 6, 2026), the Mistral Large 4 model card, and independent coverage from The New Stack, Reuters, and Local AI Zone. The open weights have not yet been released as of this writing — Mistral's timeline is "end of October 2026." All benchmark figures are vendor-reported or from cited independent evaluations; no hands-on testing of the open weights was performed.
On October 6, 2026, French AI lab Mistral AI launched Mistral Large 4 (internally nicknamed le Chonk) — a public preview of what the company describes as its largest and most capable model to date. The announcement is notable for three reasons: a trillion-parameter scale, a native multimodal architecture, and an explicit commitment to open weights by end of month.
What Mistral actually announced
- Model: Mistral Large 4 (ML4), 1 trillion total parameters, 52 billion active parameters per token.
- Architecture: Natively multimodal mixture-of-experts (MoE) — vision, text, and reasoning unified in a single model.
- Languages: Fluent in 160+ languages, including every official EU language.
- Context window: 1 million tokens.
- Training: 3,800 NVIDIA Grace Blackwell GPUs in Mistral's own European datacenters.
- Preview access: Public API on Mistral Studio and Mistral API (model ID:
mistral-large-4). - Pricing (preview): $1.36 / $4.18 per million input/output tokens. 50% launch discount through approximately October 20.
- Open weights: Committed for end of October 2026. License terms not yet published.
The benchmark claims — and the independent scores
Mistral publishes benchmark results across coding, agentic workflows, cybersecurity, finance, legal, and multimodal understanding. The most decision-relevant independent evaluation is Artificial Analysis, which has been tracking ML4 since preview launch.
| Benchmark | ML4 Score | Context |
|---|---|---|
| Cybersecurity — AA Cyber Index | Top 5 globally | Leads open weights outside China |
| Reproduce & patch (CyberGym-E2E) | 82% | Highest of any model tested |
| Cybench (40 CTF challenges) | 93% | One of highest for open weights |
| DeepSWE v1.1 | 61.7% | Ahead of DeepSeek V4 Pro, Qwen3.8 Max |
| Terminal-Bench 4.0 | 28.3% | Competitive coding agent test |
| AutomationBench (657 workflows) | 59.9% | Ahead of Kimi K3, DeepSeek V4 Pro |
| AA-Briefcase (knowledge work) | 1,393 Elo | Ahead of DeepSeek V4 Pro |
| Dense 200 (visual grounding) | 42% | Surpasses GPT-6 Astra (41%) |
Figures from Mistral's October 6 announcement and Artificial Analysis independent evaluation. Blind human evaluation (Surge AI) on coding quality ranked ML4 Preview second of five models (3.74/5), behind only Claude Opus 5 (4.22).
Why the cybersecurity angle matters
The standout claim is cybersecurity performance. Mistral argues that several leading closed models — including Claude Opus 5.5 and GPT-6 Astra — score near zero on reproduce-and-patch tests because their safety filters refuse to perform the task. Yet defending software often starts with proving a flaw is real — exactly the work those refusals block. ML4 is designed to do that work without the same constraints, and its open weights mean organisations can run it on private cloud or on-premise for sovereign security operations.
This is a structural differentiator: open weights + no refusals on defensive cyber tasks is a combination no other Western frontier lab currently offers. For organisations that need auditable, sovereign AI for security operations, this is the primary reason to track the weights release.
European sovereignty and infrastructure
ML4 was trained from scratch on Mistral's own European infrastructure — 3,800 Grace Blackwell GPUs in datacenters the company operates end-to-end under European law. The model will be available across multiple regions including a European deployment independent of other digital service providers. This matters for organisations with data sovereignty requirements, particularly in finance, defence, and public sector.
The €3 billion Series D (largest European tech equity round ever) is funding compute scale-up. Mistral states the RL run behind this preview is still in flight with no signs of saturation, implying further improvements before and after the weights drop.
What to watch
- Weights release date: Mistral says "end of October 2026." Hugging Face placeholder currently shows October 31 ETA. The exact day and license terms are not yet public.
- License terms: Mistral Large 3 is Apache 2.0. ML4 license has not been announced — verify before commercial use.
- Independent reproduction: Once weights ship, expect third-party benchmarking within days. The Cyber Index and DeepSWE claims are the ones to verify first.
- Specialised descendants: Mistral states ML4 will serve as the base for a new generation of specialised models for finance, manufacturing, cybersecurity, and other verticals.
For model evaluators: The 1T/52B MoE architecture at open weights is a new point on the Pareto frontier. Compare against Qwen3.8-2.4T-A95B (open, larger) and DeepSeek V4 Pro (open, cheaper API) on your actual workloads.
For the open ecosystem: A European lab shipping a trillion-parameter open model trained on European compute is a milestone for AI sovereignty. The weight drop timing (end of October) makes this the most imminent major open release.
Bottom line: Mistral Large 4 preview is live today; open weights are committed for end of October 2026. The cybersecurity lead among Western open weights is the strongest claim — if it holds on release, ML4 becomes the default choice for organisations that need open, sovereign, refusal-free defensive AI. Watch the license terms and independent benchmarks before committing production workloads.
- Mistral — Introducing Mistral Large 4 (official announcement, October 6, 2026)
- Mistral Docs — Mistral Large 4 model page (v26.10)
- Reuters — Mistral Large 4 coverage (October 2026)
- The New Stack — Mistral Large 4: weights by end of October
- Artificial Analysis — Mistral Large 4 provider benchmarking
- Local AI Zone — October 2026 AI Model Updates: Mistral Large 4 weights
Prices and benchmark figures are standard published rates as of October 27, 2026 and can change without notice. Some outbound links may be affiliate links — see our disclosure. Found an error? Tell us and we'll correct it in public.