Mistral

Devstral Small 1.1

Name: Devstral Small 1.1
Author: Mistral

Current

HuggingFace

37.5KDownloads363LikesJul 2025Released131K tokensContextApache 2.0License5 EntryQuality

Get started

— copy & paste to run locally

Ollama

ollama run devstral-small-2507

HuggingFace

huggingface-cli download devstral-small-2507

Quick specs

Parameters24B

Architecturedense

Context131K tokens

Modalitytext

Min RAM9.4 GB

Rec. RAM14.6 GB (Q4_K_M)

LicenseApache 2.0

FamilyDevstral Small

✓ Code✓ Reasoning

About this model

Devstral is an agentic LLM for software engineering tasks built under a collaboration between Mistral AI and All Hands AI 🙌. Devstral excels at using tools to explore codebases, editing multiple files and power software engineering agents. The model achieves remarkable performance on SWE-bench which positions it as the #1 open source model on this benchmark.

•Agentic coding: Devstral is designed to excel at agentic coding tasks, making it a great choice for software engineering agents
•lightweight: with its compact size of just 24 billion parameters, Devstral is light enough to run on a single RTX 4090 or a Mac with 32GB RAM,...
•Apache 2.0 License: Open license allowing usage and modification for both commercial and non-commercial purposes
•Context Window: A 128k context window
•Tokenizer: Utilizes a Tekken tokenizer with a 131k vocabulary size

Related models

Quick picks

Best budgetC

RX 7900 XTX 24GB~$999 — 47 tok/s

Best overallB

RTX 5090 32GB~$1,999 — 82 tok/s

Best hardware

Top picks for Devstral Small 1.1

RTX 5090 32GBB

32 GB

AMD Instinct MI100 32GBB

32 GB

RTX PRO 4500 Blackwell 32GBB

Quantization options

VRAM estimates by quant level

No hardware detected — fit column shows raw VRAM estimates

Quant	Bits	VRAM	Quality	Fit
Q2_K	2	9.4 GB	Low	—
Q3_K_S	3	11.8 GB	Low	—
NVFP4	4	13.4 GB	Medium	—
Q4_K_M	4	14.6 GB	Medium	—
Q5_K_M	5	17.3 GB	High	—
Q6_K	6	19.7 GB	High	—
Q8_0	8	25.7 GB	Very High	—
F16	16	49.2 GB	Maximum	—

Hardware compatibility

Fit estimates across all hardware

Open calculator

Computing compatibility...

Memory breakdown

Reference: NVIDIA A10 24GB

Weights14.6 GB

KV Cache3.8 GB

Runtime0.9 GB

Headroom2.4 GB

Quant

Bits

VRAM

Quality

Fit

Q2_K

9.4 GB

Low

—

Q3_K_S

11.8 GB

Low

—

NVFP4

13.4 GB

Medium

—

Q4_K_M

14.6 GB

Medium

—

Q5_K_M

17.3 GB

High

—

Q6_K

19.7 GB

High

—

Q8_0

25.7 GB

Very High

—

F16

49.2 GB

Maximum

—