Mistral Large 4

Mistral Large 4 is Mistral's largest and most capable flagship to date, positioned as a natively multimodal model built on a granular Mixture-of-Experts architecture. Mistral's documentation describes it with 52 billion active parameters and roughly 1.05 trillion total parameters, paired with a 1.6 billion parameter vision encoder that handles image understanding alongside text. The model was released as a public preview on October 6, 2026, with full open-weight availability planned for the end of that month, reflecting Mistral's strategy of pairing frontier-scale research with openly distributed checkpoints.

The model's design targets demanding professional workloads rather than casual chat, with Mistral emphasizing strong performance on coding, agentic behavior, and multimodal understanding. It is highlighted as state-of-the-art among open models for enterprise domains such as cybersecurity, finance, and law, while also exceeding closed frontier systems on visual grounding tasks. Practical fit centers on long-context analysis, with Mistral's docs listing a one-million-token context window, and on building tool-using assistants that combine reasoning, document understanding, structured outputs, and vision inputs to tackle complex real-world pipelines.

OpenCode Zenmistral-large-4mistral-large

Open provider page Provider docs

Quick Info

Provider: OpenCode Zen
Model key: mistral-large-4
Release date: Oct 6, 2026
Last updated: Oct 6, 2026
Input modalities
Output modalities
Capabilities

Cost

Input token cost: $0.68
Output token cost: $2.09

Limits

Output tokens: 262,144 tokens
Context window: 524,288 tokens

Transparent token rates

Compare Mistral Large 4 pricing

Rates are shown per one million tokens. Combined means one million input plus one million output tokens.

Browse this family

Latest news about Mistral Large 4

OpenCode Zen

CoverageBenchmark

Mistral launched a public preview of Mistral Large 4, code-named "Le Chonk" (ML4), on October 6, 2026. The model is a one-trillion-parameter multimodal system with 49 billion active parameters, trained from scratch over roughly two months on 4,000 Nvidia Grace Blackwell GPUs in Mistral's European data centers. ML4 targets coding, cybersecurity, finance, manufacturing, and visual grounding, with training across more than 160 languages including every official EU language. Mistral plans to publish the model weights on October 27 under a custom license following a three-week developer and government testing period, positioning the release as an open-weight foundation for sovereign and enterprise deployments.

OpenCode Zen

CoverageBenchmark

LLM Stats lists Mistral Large 4 with a blended price of $0.75 per million tokens and an LLM Stats Score of 46.4, placing it between GPT OSS 120B and Claude Haiku 5.5 on cost efficiency. The tracker surfaces several benchmark results attributed to mistral.ai's announcement, including CyBench at 0.93, SciCode at 0.92, and CyberGym at 0.82. The tracker also reports an AA-Briefcase Elo of 1393 out of 3000 and flags the benchmark figures as self-reported by the model provider rather than independently re-run. These results support the model's positioning toward agentic cybersecurity and scientific coding workloads highlighted in the launch.

OpenCode Zen

CoverageBenchmark

BenchLM reports Mistral Large 4 with a capability score of 53.7/100, ranking 69 of 214 tracked models as of October 6, 2026. The model is listed with a 1M-token context window, 116 tok/s output speed, and a first-token latency of 18.69 seconds, indicating long-context inference optimization. Pricing on BenchLM is recorded at $0.68 input and $2.09 output per million tokens, with a $0.07 cached input rate. The page shows only 3 published benchmark rows across agentic, coding, and reasoning categories, with most other capability areas marked as "Not measured," signaling thin independent benchmark coverage at launch.

OpenCode Zen

CoverageBenchmark

Vals AI tracks Mistral Large 4 as an open-weight model released October 6, 2026, with a 512k context window, 256k max output tokens, and support for text and image input. The model ranks 32 of 44 on the Vals Index with a composite score of 48.05%, with its best result being 6 of 75 on Harvey's Legal Agent Benchmark. Token pricing on the tracker is listed at $1.36 input / $4.18 output per million, with a cost-per-test of $13.78 and a measured latency of 105 minutes 50 seconds. Vals AI notes ML4 shows particular strength on legal and finance agent benchmarks relative to its overall ranking position.