Mistral AI

Mistral Small 4

Name: Mistral AI Mistral Small 4
Author: Mistral AI

Try It Compare

Model ID:mistralai/mistral-small-2603

2026-03-16

Try It Compare

Mistral Small 4 is the first Mistral model to unify three flagship capabilities into a single system: strong reasoning from Magistral, multimodal understanding from Pixtral, and agentic coding from Devstral. Built on a 119B-parameter Mixture-of-Experts architecture activating just 6.5B parameters per token, it offers a 256K-token context window with configurable reasoning effort — from lightweight instant responses to deep step-by-step analysis. Fully open source, it delivers a 40% reduction in latency and 3× throughput improvement over Mistral Small 3, making it a versatile and efficient choice for coding, analysis, and vision tasks.

API|VisionReasoning|Open ModelApache 2.0

Knowledge Cutoff

2025

The date this AI finished learning. It may not know about things that happened after this date.

Input → Output Format

The types of content this AI can receive, and what it can produce in return.

Context Memory

262KIN262KOUT

The maximum amount of text the AI can read and process in a single request. A larger number means it can handle longer documents or conversations.

Cost/1M Words

$0.15IN$0.6OUT

The cost of using this AI directly in your own application. Shown in USD per 1 million units of text (tokens).

Calculate Cost

Source:Official Docs OpenRouter

AI Performance Evaluation

Overall

AA Intelligence Index

28%↓11%

Reasoning & Math

GPQA Diamond

77%↓4%

HLE

9.5%↓8%

MMLU-Pro

42%↓40%

MATH-500

56%↓37%

Coding

AA Coding Index

24%↓10%

LiveCodeBench

11%↓54%

TAU2

41%↓32%

TerminalBench

17%↓14%

SciCode

38%↓3%

Language & Instructions

IFBench

48%↓9%

AA-LCR

45%↓17%

Output Speed

Standard Mode

151tok/s↑69

First Output 0.52s

Reasoning Mode

164tok/s↑76

First Output 12.79s

Source:Artificial Analysis

Mistral AI