Mistral releases Large 4, a 1-trillion-parameter model, in public preview

Mistral AI published Mistral Large 4 on 6 October 2026. Quote: "Introducing Mistral Large 4"

Introducing Mistral Large 4 Back to Blog 14 min read October 6, 2026 By Mistral

Mistral describes Mistral Large 4 as a 1 trillion-parameter natively multimodal model with 49 billion active parameters. Quote: "ML4 is a 1 trillion-parameter natively multimodal model with 49 billion active parameters."

ML4 is a 1 trillion-parameter natively multimodal model with 49 billion active parameters.

Mistral lists the preview API price as $1.36 per million input tokens and $4.18 per million output tokens. Quote: "Input (/M tokens) $1.36 Output (/M tokens) $4.18"

Input (/M tokens) $1.36 Output (/M tokens) $4.18

Mistral says it will release the weights by the end of the month. Quote: "We will release the weights by the end of the month."

We will release the weights by the end of the month.

Until the weights are released, Mistral says it is red-teaming the model with cybersecurity leaders, vetted partners and state authorities, who access the same model with reduced moderation and expanded cyber capabilities. Quote: "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."

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.

Mistral says Large 4 was trained from scratch on 3,800 NVIDIA Grace Blackwell GPUs in its own datacenters in Europe, and the public preview is served on the same infrastructure. Quote: "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."

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.

Mistral says that on the Artificial Analysis Cyber Index Large 4 ranks among the top five models globally; this is Mistral's own account of the result. Quote: "it ranks among the top five models globally and leads open-weight models developed outside China by a wide margin."

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.

Mistral reports Large 4 scores 61.7% on DeepSWE v1.1, 59.4% on SWE-Atlas-QnA and 28.3% on Terminal-Bench 4. Quote: "scoring 61.7% on DeepSWE v1.1, 59.4% on SWE-Atlas-QnA, and 28.3% on Terminal-Bench 4."

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.

Mistral says that on Lakera's public B3 AI Security Benchmark Large 4 resists 93.3% of attacks. Quote: "On Lakera’s public B3 AI Security Benchmark , ML4 resists 93.3% of attacks"

On Lakera’s public B3 AI Security Benchmark , ML4 resists 93.3% of attacks

Mistral says a significant share of Large 4's training data was multilingual, spanning more than 160 languages, including every official language of the European Union. Quote: "a significant share of ML4’s training data was multilingual, spanning more than 160 languages, including every official language of the European Union."

a significant share of ML4’s training data was multilingual, spanning more than 160 languages, including every official language of the European Union.

Mistral describes Large 4 as an open-weight hybrid instruct-and-reasoning MoE with multimodal input. Quote: "Open-weight hybrid instruct-and-reasoning MoE with multimodal input"

Open-weight hybrid instruct-and-reasoning MoE with multimodal input