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

Mistral AI - Dense - verified

Run Devstral 2 123B Instruct 2512 locally

Devstral 2 123B Instruct 2512 is a 125B dense model from Mistral AI with a 262K-token context window, released under Modified MIT License (commercial use is allowed with conditions set out in the license); the smallest listed download is 70.3 GB, the 4bit MLX build from mlx-community/Devstral-2-123B-Instruct-2512-4bit.

Official weights: mistralai/Devstral-2-123B-Instruct-2512

Total params
125B
Architecture
Dense
Context
262K
Smallest download
70.3 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.

Devstral 2 123B Instruct 2512 downloads and file sizes
DownloadFormatSizeRepo
Official weightsFP8128.2 GBmistralai/Devstral-2-123B-Instruct-2512
Q4_K_MGGUF74.9 GBunsloth/Devstral-2-123B-Instruct-2512-GGUF
Q8_0GGUF132.9 GBunsloth/Devstral-2-123B-Instruct-2512-GGUF
4bitSmallest listedMLX70.3 GBmlx-community/Devstral-2-123B-Instruct-2512-4bit

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.

Ollama

Terminal
ollama run devstral-2

Source: ollama.com/library/devstral-2

vLLM

Terminal
vllm serve mistralai/Devstral-2-123B-Instruct-2512 \
    --tool-call-parser mistral --enable-auto-tool-choice \
    --tensor-parallel-size 8

Source: huggingface.co/mistralai/Devstral-2-123B-Instruct-2512

SGLang

Terminal
python -m sglang.launch_server --model-path mistralai/Devstral-2-123B-Instruct-2512 --host 0.0.0.0 --port 30000 --tp 8 --tool-call-parser mistral

Source: huggingface.co/mistralai/Devstral-2-123B-Instruct-2512

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

License

Commercial use
Commercial use: ConditionsCommercial use is allowed with conditions set out in the license.
Gated download
No. The files download directly.

Worth knowing

Devstral 2 ships under a modified MIT license that excludes companies with global monthly revenue above 20 million USD.

Released (per the announcement).

FAQ

How big is the Devstral 2 123B Instruct 2512 download?

The official FP8 weights in mistralai/Devstral-2-123B-Instruct-2512 are 128.2 GB. The smallest listed download is 70.3 GB, the 4bit MLX build from mlx-community/Devstral-2-123B-Instruct-2512-4bit. Plan for more memory than the file size: the context (KV cache) and the operating system need room too.

Can I use Devstral 2 123B Instruct 2512 commercially?

Devstral 2 123B Instruct 2512 is released under Modified MIT License, and commercial use is allowed with conditions set out in the license. The license text is at https://huggingface.co/mistralai/Devstral-2-123B-Instruct-2512/blob/main/LICENSE.

How much memory do I need for Devstral 2 123B Instruct 2512?

By this directory's rule of thumb it fits in 128 GB and up: the smallest listed download, 70.3 GB, is at most 85% of 128 GB, which leaves room for the context and the operating system.

How do I run Devstral 2 123B Instruct 2512 with Ollama?

Run "ollama run devstral-2", as shown at https://ollama.com/library/devstral-2. The Ollama library lists the devstral-2:123b tag at 75GB.

What is the context length of Devstral 2 123B Instruct 2512?

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.