[whatmodel.sh](../../README.md) / Devstral 2 (2512)
# Devstral 2 (2512)
Mistral · released 2025-12 · 123B params · modified-mit
**score 61.5** ±1.6 · #191 of 732 · from 5 benchmark results across 3 sources
Score within the [LMArena (text)](https://whatmodel.sh/?set=lmarena-text&score=set) comparison set only.
## What it's good at
● marks Devstral 2 (2512) among every model on that task or test, worst to best.
task vs the others rank counts tests
agent tasks ▁▁▃▆█▄▅●▄▄▂▁ #112 of 313 34% 3
coding ▁▁▁▃▅▇█●▅▄▃▁ #194 of 469 70% 3
agent tasks · #112 of 313 · 34% of its score
Beats [Grok 4.20 (0309 reasoning)](grok-4-20.md), [o3-pro](o3-pro.md), [Grok 4.3](grok-4-3.md) here, though they rank higher overall. Trails [SmolLM3 3B](smollm3-3b.md) here, though they rank lower overall.
test result vs the others in its run difficulty counts why
SWE-bench Verified (maker) 72.2% ▁▁▁▁▁▁▂▄●▆▁▁ beats 90/138 hard · avg 33% 15% sharp · at its level · small run
Terminal-Bench 2.0 (maker) 32.6% ▂▂▅●████▆▂▃▃ beats 7/36 very hard · avg 11% 11% at its level · small run
SWE-bench Multilingual (maker) 61.3% ▁··▁▃▁▅●█▂▂▁ beats 23/65 hard · avg 30% 9.3% at its level · small run
coding · #194 of 469 · 70% of its score
Beats [Gemini 2.5 Flash](gemini-2-5-flash.md) here, though they rank higher overall. Trails [QwQ-32B-Preview](qwq-32b-preview.md), [Ministral 3 8B Reasoning 2512](ministral-3-8b-reasoning.md), [OpenThinker 32B](openthinker-32b.md) here, though they rank lower overall.
test result vs the others in its run difficulty counts why
Arena WebDev (Code Arena) 1197.1 ▁▁●▂▅█▃▆▅▅▁▂ beats 5/129 avg 981 46% sharp
SWE-bench Verified (maker) 72.2% ▁▁▁▁▁▁▂▄●▆▁▁ beats 90/138 hard · avg 33% 15% sharp · at its level · small run
SWE-bench Multilingual (maker) 61.3% ▁··▁▃▁▅●█▂▂▁ beats 23/65 hard · avg 30% 9.3% at its level · small run
other tests (1)
test result vs the others in its run difficulty counts why
DSEffi-Bench Pass (1,000 data-sci… 43.0% ▅··▃▅·●·▃██▃ beats 7/15 hard · avg 23% 18% at its level
source results share of the score
LMArena leaderboard dataset 1 46% ▇▇▇▇▇▇
mistralai/Devstral-Small-2-24B-Instruct… 3 34% ▇▇▇▇
DSEffi-Bench: Demystifying Large Langua… 1 18% ▇▇
prior: pull toward the middle 1% ▇
weighted average = 61.5
Licences: LMArena leaderboard dataset ([CC BY 4.0](https://creativecommons.org/licenses/by/4.0/)) · mistralai/Devstral-Small-2-24B-Instruct-2512 model card ([Apache-2.0](../../licenses/cards-gap-NOTICE.txt)) · DSEffi-Bench: Demystifying Large Language Models' Capability in Efficient Data Science Code Generation ([CC BY 4.0](../../licenses/paper-2608.30248v1-CC-BY-4.0.txt))
## How to run it
### llama.cpp (recommended)
1. Install llama.cpp. Homebrew and winget installs name the commands llama-cli and llama-server instead. [Other ways to install](https://github.com/ggml-org/llama.cpp/blob/master/docs/install.md)
```sh
# macOS / Linux
curl -LsSf https://llama.app/install.sh | sh
# Windows
irm https://llama.app/install.ps1 | iex
```
2. Run Devstral 2 (2512) (Q4_K_M, needs ~74.9 GB, from [huggingface.co/lmstudio-community](https://huggingface.co/lmstudio-community/Devstral-2-123B-Instruct-2512-GGUF/tree/330d61a4b7ed244a64a6bc3f7e98b250c3264546))
```sh
llama cli -hf lmstudio-community/Devstral-2-123B-Instruct-2512-GGUF:Q4_K_M
```
3. Or start a local server and open the address it prints (http://localhost:9931)
```sh
llama serve -hf lmstudio-community/Devstral-2-123B-Instruct-2512-GGUF:Q4_K_M
```
Other versions: Q8_0 (~132.9 GB) `llama cli -hf lmstudio-community/Devstral-2-123B-Instruct-2512-GGUF:Q8_0`
### LM Studio
1. Install LM Studio. Open the app once before using the lms command. [Download it (macOS / Windows / Linux)](https://lmstudio.ai/download)
2. Run Devstral 2 (2512) (Q4_K_M, needs ~74.9 GB, from [huggingface.co/lmstudio-community](https://huggingface.co/lmstudio-community/Devstral-2-123B-Instruct-2512-GGUF/tree/330d61a4b7ed244a64a6bc3f7e98b250c3264546))
```sh
lms get https://huggingface.co/lmstudio-community/Devstral-2-123B-Instruct-2512-GGUF@Q4_K_M -y
```
3. Then run lms ls to see its name, and lms chat followed by that name
Other versions: Q8_0 (~132.9 GB) `lms get https://huggingface.co/lmstudio-community/Devstral-2-123B-Instruct-2512-GGUF@Q8_0 -y` · 4bit (~70.3 GB, Mac only) `lms get https://huggingface.co/mlx-community/Devstral-2-123B-Instruct-2512-4bit -y` · 8bit (~132.8 GB, Mac only) `lms get https://huggingface.co/mlx-community/Devstral-2-123B-Instruct-2512-8bit -y`
### Ollama
1. Install Ollama. [Download it (macOS / Windows)](https://ollama.com/download)
```sh
# Linux
curl -fsSL https://ollama.com/install.sh | sh
```
2. Run Devstral 2 (2512) (Q4_K_M, needs ~74.3 GB, from [ollama.com](https://ollama.com/library/devstral-2/tags))
```sh
ollama run devstral-2:123b-instruct-2512-q4_K_M
```
Other versions: Q8_0 (~148.1 GB) `ollama run devstral-2:123b-instruct-2512-q8_0`
### MLX (Mac)
1. Install MLX (Mac). Apple silicon Macs only, text only.
```sh
# with uv
uv tool install mlx-lm
# or
pip install mlx-lm
```
2. Run Devstral 2 (2512) (4bit, needs ~70.3 GB, Mac only, from [huggingface.co/mlx-community](https://huggingface.co/mlx-community/Devstral-2-123B-Instruct-2512-4bit/tree/4df017152f869b334922f9dd7a7e8d5dca42cf0b))
```sh
mlx_lm.chat --model mlx-community/Devstral-2-123B-Instruct-2512-4bit
```
3. Or start an OpenAI-compatible server at http://localhost:8080/v1
```sh
mlx_lm.server --model mlx-community/Devstral-2-123B-Instruct-2512-4bit
```
Other versions: 8bit (~132.8 GB, Mac only) `mlx_lm.chat --model mlx-community/Devstral-2-123B-Instruct-2512-8bit`
Model repository and documentation: [huggingface.co/mistralai](https://huggingface.co/mistralai/Devstral-2-123B-Instruct-2512)
[interactive version](https://whatmodel.sh/model/devstral-2) · [JSON](https://whatmodel.sh/api?format=json&all=1)