Mistral AI · Code Models

Devstral Small

Name: Devstral Small
Author: Mistral AI

An agentic coding model by Mistral AI and All Hands AI, built to solve real software engineering tasks inside coding agents, runnable on one GPU.

Official page

Run it locally

With Ollama installed, download and start the default version with:

ollama run devstral:24b

Variant	Download
devstral:24b	14GB

Memory: About 14GB for the default devstral:24b version (4-bit)
Formats: Safetensors (Hugging Face), GGUF (Ollama, llama.cpp, LM Studio)

The download size is roughly the memory (RAM or GPU VRAM) needed to load a variant; allow extra for long contexts. See our hardware guide.

Overview

Devstral was developed by Mistral AI with All Hands AI and fine-tuned from Mistral Small for working in agent frameworks that edit repositories and run tests.

Highlights

24B parameters
Designed for coding agents
128K-token context window
Apache 2.0 license

Typical uses

Local coding agents, repository-level code changes and automation.

Things to consider

Text only.
Works best inside an agent scaffold.
Download sizes refer to the default 4-bit quantized versions in the Ollama library; full-precision weights are larger.

Specifications are taken from the model card on Hugging Face, the Ollama library and the provider’s announcement. Benchmark results depend on evaluation settings; see how AI benchmarks work.

#Code Models #Open Weights #Local AI #Mistral AI