whatmodel.sh / Mistral Large 4 (October 29 weights)
Mistral Large 4 (October 29 weights)
↗ share
+ contribute
Mistral Large 4 (October 29 weights) is an open-weights language model from Mistral, released October 2026. Its published size is 1000B parameters (52B active per token). Licence: Mistral Ecosystem Open-Weight License.
Catalogue identity notes
Improved October 29 open-weight release following three weeks of additional reinforcement learning, distinct from the October 6 public preview. Preview benchmark results are not assigned to this checkpoint. No verified eligible exact-release benchmark imported yet; unknown API price omitted. Custom ecosystem licence allows commercial/noncommercial use subject to conditions; Model-as-a-Service businesses require a separate commercial licence. Native text/image inputs, reasoning high by default; NVFP4 is a separate artifact and its quality is not assumed identical.
Model information sources
Release date ↗ · Weights licence ↗ · Parameter count ↗
Maker describes approximately 1 trillion total and 52B active parameters including embeddings/output layers (49B without them); nominal size, not a summed weight inventory.
$ how it compares
Scored LLMs released closest to Mistral Large 4 (October 29 weights), out of 1194 with a score, then more from Mistral.
| # | model | score | where | ||
|---|---|---|---|---|---|
| 3 |  GPT-6.1 Terra | 84±1 | api $2.00/M | ||
| 4 |  Gemini 4 Carbon | 83±1 | limited preview | ||
| 7 |  Muse Wonder Full | 82±1 | Muse Code only | ||
| 16 |  Kimi K3.1 Pro | 79±2 | api price unknown | ||
| 18 | Muse Wonder | 78±1 | api $8.00/M | ||
| more from Mistral | |||||
| 83 | Mistral Large 4 (public preview) | 71±1 | comparison only | ||
| 207 | Mistral Medium 3.5 (128B) | 64±1 | api $3.00/M +local | ||
| 213 | Mistral Medium 3.1 (2508) | 64±2 | comparison only | ||
$ where to run it
Through an API
api · Mistral AI · price unknown · docs.mistral.ai ↗
Other ways to run it
- Run it as published · needs ~2000 GB, NVIDIA only · from huggingface.co/mistralai
docker run --runtime nvidia --gpus all -v ~/.cache/huggingface:/root/.cache/huggingface -p 8000:8000 --ipc=host vllm/vllm-openai:mistral-large-4 --model mistralai/Mistral-Large-4-1T-A52B --tensor-parallel-size 8 --tool-call-parser mistral --reasoning-parser mistral --enable-auto-tool-choice --gpu-memory-utilization 0.85
Model repository and documentation: huggingface.co/mistralai ↗
$ improve this page
Use a short GitHub form or edit the data through a pull request. Every number needs a source; here's which ones we can use.
data updated · sources and licences · json