Modelpeers
Mistral Large 4 character

ModelMistral AI · Joined Oct 6, 2026

Mistral Large 4

Frontier Model at Mistral AI | Open-weight multimodal model in public preview | 1M-token context

Available for Reasoning and coding and cybersecurity

Compare
Official pageReport a correction

The basics

Checked Oct 22
Status
Public previewIn public preview on the Mistral API since Oct 6, 2026. Mistral says open weights come by the end of October.
Model ID
Price
$1.36 in, $4.18 outPer 1M tokens, changed Oct 6
Context
1M tokens

Updates

Full changelog, by month

Strengths and weaknesses

Strengths

  • Reasoning

  • Coding

  • Image understanding

  • Cybersecurity

Weaknesses

  • !

    Weights are not downloadable yet and no license is published

  • !

    Some providers serve a 524K-token context, not 1M

Role

Flagship model at Mistral AI

  1. Mistral Large 4 character

    Mistral Large 4Current

    Oct 2026 to now · 1 month

Works with

From recorded changes
Microsoft Foundry

API · 1 shared change

Mistral AI Studio

API · 1 shared change

Similar models

GPT-6 Astra character
GPT-6 Astra

$10 in, $50 out

Claude Opus 5 character
Claude Opus 5

$5 in, $25 out

Muse Spark 1.3 character
Muse Spark 1.3

$1.25 in, $4.25 out

Similar models for now. Real viewing data comes after launch.

Cost vs peers

Cheapest of 5 similar models by output price.

Standard API rates for uncached input on the shortest prompts, from each maker's pricing page. Gemini 4 Argon: Introductory price through Nov 30, 2026, then $4 in and $20 out.

Cost and specs

Cost

Per 1M tokens

Input
$1.36
Output
$4.18
Cached input
$0.14

Specs

Context
1M tokens
Parameters
1.05T total, 52B active
Vision encoder
1.6B parameters
Version
v26.10
Max output
262,144 tokens
Input
Text, images and PDFs

Where it works

PlacePrice per 1M tokens, in / outUptime
Mistral APImistral-large-4
OpenRoutermistralai/mistral-large-4-0
Vercel AI Gatewaymistral/mistral-large-4

Benchmarks

Higher is better. Green is Mistral Large 4. Dark marks are similar models.

Independent

Self-reported by Mistral AI

Terminal-Bench 4.0

Command-line agent tasks

28.3%

Cybench

Capture-the-flag security tasks

93%

DeepSWE v1.1

Software engineering tasks

61.7%

What people say

Reviews

No reviews yet. Be the first to review Mistral Large 4.

Used Mistral Large 4? Tell other builders what worked and what did not.

Research sources

Researched Oct 22, 2026. Each section's own sources open with its Show sources button.