Skip to main content
Watch: I Built an AI App Without Leaving Databricks

Mistral AI - Mixture of experts - verified

Run Mistral Small 4 119B (2603) locally

Mistral Small 4 119B (2603) is a 119.4B mixture-of-experts (6.5B active) model from Mistral AI with a 262K-token context window, released under Apache 2.0 (commercial use is allowed); the smallest listed download is 67.8 GB, the 4bit MLX build from mlx-community/Mistral-Small-4-119B-2603-4bit.

Official weights: mistralai/Mistral-Small-4-119B-2603 - modalities: text, image

Total params
119.4B
Active params
6.5B
Context
262K
Smallest download
67.8 GB

Fits in

Smallest download at most 85% of memory

Downloads

The official weights and every quantization listed for this model, with file sizes from Hugging Face.

Mistral Small 4 119B (2603) downloads and file sizes
DownloadFormatSizeRepo
Official weightsFP8120.9 GBmistralai/Mistral-Small-4-119B-2603
UD-Q4_K_MGGUF73.8 GBunsloth/Mistral-Small-4-119B-2603-GGUF
Q8_0GGUF126.5 GBunsloth/Mistral-Small-4-119B-2603-GGUF
4bitSmallest listedMLX67.8 GBmlx-community/Mistral-Small-4-119B-2603-4bit

GB is decimal: 1 GB = 10^9 bytes. Add memory for the context and the operating system on top of the file size.

Run it

Copied verbatim from the page linked under each command. Server commands from model cards often assume a multi-GPU node; check the flags before you run them.

vLLM

Terminal
vllm serve mistralai/Mistral-Small-4-119B-2603 --max-model-len 262144 --tensor-parallel-size 2 --attention-backend FLASH_ATTN_MLA \
  --tool-call-parser mistral --enable-auto-tool-choice --reasoning-parser mistral --max_num_batched_tokens 16384 --max_num_seqs 128 \
  --gpu_memory_utilization 0.8

Source: huggingface.co/mistralai/Mistral-Small-4-119B-2603

Not sure which runtime to use? Compare local LLM runtimes by platform, API and features.

License

License
Apache 2.0apache-2.0
Commercial use
Commercial use: YesCommercial use is allowed.
Gated download
No. The files download directly.

Worth knowing

Mistral Small 4 is a 128-expert MoE with 4 experts active per token that unifies the Instruct, Reasoning (Magistral) and Devstral lines in one model.

Released (per the announcement).

API pricing and specsHosted prices per 1M tokens for Mistral Small 4 119B (2603), if you would rather not run it yourself.

FAQ

How big is the Mistral Small 4 119B (2603) download?

The official FP8 weights in mistralai/Mistral-Small-4-119B-2603 are 120.9 GB. The smallest listed download is 67.8 GB, the 4bit MLX build from mlx-community/Mistral-Small-4-119B-2603-4bit. Plan for more memory than the file size: the context (KV cache) and the operating system need room too.

Can I use Mistral Small 4 119B (2603) commercially?

Mistral Small 4 119B (2603) is released under Apache 2.0, and commercial use is allowed. The license text is at https://www.apache.org/licenses/LICENSE-2.0.

How much memory do I need for Mistral Small 4 119B (2603)?

By this directory's rule of thumb it fits in 128 GB and up: the smallest listed download, 67.8 GB, is at most 85% of 128 GB, which leaves room for the context and the operating system.

What is the context length of Mistral Small 4 119B (2603)?

262,144 tokens. A longer context needs more memory for the KV cache, so you can run with a shorter one when memory is tight.

Sources

Last verified from the pages below.

Newsletter

New open weights, every week

One email per week. New model releases, open-weight drops and what they mean for what you build.

Keep exploring

Best local AI models

Our curated picks by hardware tier, with benchmarks and the reasoning.

Model API pricing and specs

First-party prices per 1M tokens when you would rather call an API.

AI benchmark leaderboards

Sourced scores, for when you need to compare quality.

Setup guides

Point your coding tools at the model you picked.