Mistral Large 4, nicknamed Le Chonk, is a natively multimodal mixture-of-experts model from Mistral that accepts interleaved text and image input and produces text output. It uses a granular MoE design with roughly 1.05 trillion total parameters and 49 billion active per token, reported as 52 billion when embeddings and output layers are counted, paired with a 1.6 billion parameter vision encoder and a context window of one million tokens. The model is trained from scratch on about 3,800 NVIDIA Grace Blackwell GPUs in Mistral's European data centers and supports more than 160 languages. It behaves as a hybrid instruct and reasoning system, with a reasoning effort setting that selects between direct answers and longer deliberation, alongside function calling and structured output for agentic workflows.
Image understanding is a focus of this generation, covering documents, charts, technical drawings and natural scenes, and the model emits bounding box coordinates for visual grounding queries. Reported grounding results include 42 percent on Dense200 and 73 percent on the DIOR-RSVG remote sensing benchmark. Mistral describes agentic vision workflows in which the model zooms into gigapixel satellite imagery or engineering drawings to verify details, and reports coding results such as 62 percent on DeepSWE. Figures published at preview time are preliminary because the reinforcement learning phase is still in progress.
Drag and drop an image here, or click to browse
Model settings
Thinking level
Max output tokens
Default 65,536 · max 65,536
Sign in to adjust thinking and output length per run.
Results appear here. Add an image or pick an example to run Mistral Large 4.
—
Usage
Past 30 DaysVision Evals is Roboflow's ground-truth benchmark: every model runs the same real-world samples across six vision tasks, and answers are scored against ground truth.
Evals updated October 8, 2026Pricing updated October 8, 2026
Mistral Large 4 averages 68.5% across the six Vision Evals tasks, ranking #36 of 61 models overall.
Its weakest relative showing is Reasoning, ranking #47 of 61 at 38.9%.
At $0.0018 per sample it is the 26th cheapest of the 61 benchmarked models, and its average inference time of 8.8s per sample makes it the 27th fastest.
Field medians: Object Detection 54.3%, Counting 62.6%, Identification 84.4%, OCR 89.4%, Data Extraction 84.5%, Reasoning 57.6%.
| Task | Score | Field (0 to 100) | Rank | Cost / sample | Speed |
|---|---|---|---|---|---|
| Object Detection (low) | 59.3% ±0.7, Mean of 3 runs, range 58.5 to 60.0 | #22 of 61 | $0.0028 | 18.44s | |
| Object Detection (high) | 50.2% ±2.5, Mean of 3 runs, range 48.0 to 53.0 | #22 of 29 | $0.023 | 172.22s | |
| Counting (low) | 54.5% ±0.7, Mean of 3 runs, range 54.0 to 55.4 | #40 of 61 | $0.0010 | 1.61s | |
| Counting (high) | 63.1% ±2.0, Mean of 3 runs, range 60.8 to 64.9 | #24 of 29 | $0.0094 | 58.46s | |
| Identification (low) | 85.4% ±3.1, Mean of 3 runs, range 81.3 to 87.5 | #27 of 61 | $0.0009 | 1.50s | |
| Identification (high) | 85.4% ±1.6, Mean of 3 runs, range 84.4 to 87.5 | #22 of 29 | $0.0044 | 26.22s | |
| OCR (low) | 92.7% ±0.8, Mean of 3 runs, range 91.8 to 93.3 | #8 of 61 | $0.0016 | 5.20s | |
| OCR (high) | 87.1% ±4.4, Mean of 3 runs, range 81.6 to 90.4 | #27 of 29 | $0.025 | 165.33s | |
| Data Extraction (low) | 80.1% ±1.0, Mean of 3 runs, range 79.4 to 81.4 | #46 of 61 | $0.0010 | 1.68s | |
| Data Extraction (high) | 82.5% ±1.0, Mean of 3 runs, range 81.4 to 83.5 | #22 of 29 | $0.0033 | 16.08s | |
| Reasoning (low) | 38.9% ±0.3, Mean of 3 runs, range 38.4 to 39.1 | #47 of 61 | $0.0013 | 3.27s | |
| Reasoning (high) | 57.6% ±2.0, Mean of 3 runs, range 55.6 to 59.6 | #41 of 48 | $0.013 | 91.80s |
Overall benchmark score against estimated cost per sample, on a log scale. Upper-left is the sweet spot: high quality at low cost.
60 models on the current benchmark · scores and efficiency pooled across all six tasks at low effort · Mistral Large 4 highlighted
Mistral Large 4 scores are the mean of 3 runs per task at both low and high effort · Methodology
View all Vision Evals →Mistral Large 4 costs $0.680 per 1M input tokens and $2.09 per 1M output tokens.
Pricing updated Oct 8, 2026
Mistral Large 4 ships under a custom, model-specific license rather than a standard permissive or restrictive one, so the Mistral Large 4 license has to be read directly. Custom model licenses range from effectively permissive to research-only.
Uncertainty around licensing can delay or stop a project, and acceptable-use policies attached to custom licenses are binding terms rather than guidance. Review them alongside the Mistral Large 4 license before production deployment.
If the custom terms rule out your use case, a commercial license from the rights holder is the way through. Roboflow's licensing page lists the supported models whose commercial license is included in a Roboflow plan, so it is worth checking whether Mistral Large 4 — or a permissively licensed alternative — fits your deployment.
Do not hesitate to reach out with questions for your commercial project — our team will help you start solving business problems on the first call. See Roboflow commercial licensing for the models included in each plan.
Talk to salesThis model is released under a custom license that does not match a standard open-source identifier. Read the full license text linked from the model documentation.
Custom licenses vary widely in what they permit. Many model-specific custom licenses include commercial-use restrictions (e.g., non-commercial weights, named-user limits, or jurisdiction restrictions). Read the full license before deploying commercially.
Custom licenses are model-specific. Always check the per-model License Notes section above and the linked official license text.
License information is provided as a guide and is not legal advice.
Yes. Mistral Large 4 accepts image input and handles OCR, data extraction, object counting, identification, visual reasoning, and object detection. On Roboflow's Vision Evals its strongest task is OCR at 92.7% (#8 of 61 at low effort). You can test it on your own image in the demo above.
Yes, and it is one of the model's strongest vision skills: its transcriptions match the ground truth 92.7% on average (#8 of 61 at low effort) on Vision Evals OCR. Pulling specific fields out of documents (data extraction) scores 80.1%.
Not its strength. On Vision Evals, Mistral Large 4 scores 59.3% mAP@50 on object detection (#22 of 61 at low effort) and 54.5% judge-graded accuracy on object counting. For production counting or precise localization, pairing it with a specialized detector like RF-DETR or your own trained model in a Roboflow Workflow is usually more reliable: detect the objects, then count the detections.
On our benchmark's task mix, Mistral Large 4 averages $0.0018 per sample at $0.68 per 1M input and $2.09 per 1M output tokens (#26 of 61 on cost), with an average speed of 8.8s per sample across the benchmark. Actual cost depends on your images and prompts.
On the overall Vision Evals ranking, Mistral Large 4 sits #36 of 61 at 68.5%, just behind Claude Opus 4.8 (68.7%) and just ahead of Qwen3.7 Plus (67.4%). See the full side-by-side: Mistral Large 4 vs Claude Opus 4.8.