** It is available today through the public preview API on Mistral Studio, and the full weights will drop at month-end.
Key Capabilities
- Weight integration plan – After the end-of-month release, you can host the model on NVIDIA Grace Blackwell GPUs for parity with training performance and full control over updates.
Overview of Mistral Large 4
Therefore, it delivers state-of-the-art results on coding, agentic workflows, and multimodal understanding. Consequently, enterprises in cybersecurity, finance, and law report it as the leading open model for their workloads.
Training Infrastructure and Multilingual Data
Additionally, the public preview runs on this same hardware, guaranteeing consistent latency and throughput.
Model Architecture and Parameter Composition
The architecture builds on transformer blocks optimized for both dense and sparse attention, enabling efficient scaling across the massive parameter count.
Deployment, Access, and Weights Release Schedule
- Deployment: Use the hosted API for immediate inference or export the released weights for self-hosting.
- Weights release: End of the current month; after release, you can deploy on NVIDIA Grace Blackwell GPUs to match training performance.
Therefore, organizations that require on-premise control should prepare a secure environment now to avoid downtime when the weights become available.
Cybersecurity Strengths
In benchmark tests, it:
Thus, it serves as a reliable component in automated vulnerability-remediation pipelines.
Safety and Red-Team Practices
It also exhibits the highest refusal rate on malicious cyber prompts among open models. **
Weight Integration and Self-Hosting
To maintain performance parity after the weight drop:
- Provision NVIDIA Grace Blackwell GPUs matching the training cluster.
- Configure a secure, isolated runtime environment with encrypted storage.
- Deploy the released weights using Mistral’s containerized inference stack.
Furthermore, this setup grants full control over model updates and compliance with internal data-governance 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. 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.
Red-Teaming, Moderation, and Safety Practices
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).
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.
Performance Benchmarks
Cybersecurity Capabilities
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.
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. 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 Performance
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 Workflow Efficiency
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. On AA-Briefcase, which evaluates long-horizon knowledge work, it reaches 1,393 Elo, ahead of DeepSeek V4 Pro.
Multimodal Understanding and Visual Grounding
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%).
Scientific and Mathematical Reasoning
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. 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 with more structure than GLM-5.3, and it can sustain long, domain-specific applied-mathematics tasks, including work relevant to frontier theoretical physics.
Knowledge Work Proficiency
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.
On HarveyAI’s Legal Agent benchmark, ML4 outperforms all open-source models.
Comparative Positioning with Other Leading Models
| Benchmark | ML4 | Closest Competitor |
|---|---|---|
| Dense 200 (visual grounding) | 42% | GPT-6-Astra 41% |
| Cybench success rate | 93% | Claude Opus 5.5 near 0% |
| DeepSWE v1.1 coding | 61.7% | DeepSeek V4 Pro lower |
| AutomationBench workflow | 59.9% | Kimi K3 lower |
| AA-Briefcase Elo | 1,393 | DeepSeek V4 Pro lower |
| SciCode-Verified | State-of-the-art (open-weight) | GLM-5.3 lower |
Future Outlook and Model Extensions
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.
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. 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.