---
title: "Mistral Large 4 enters preview with 1 trillion parameters, open weights due this month"
url: https://anawords.com/story/mistral-large-4-public-preview
published: 2026-10-11T19:47:50+05:30
updated: 2026-10-11T19:47:50+05:30
section: Models & Launches
author: Mayank Sahu
checked: sources checked by AI, approved by editor
version: 1
---

# Mistral Large 4 enters preview with 1 trillion parameters, open weights due this month

The French lab’s biggest model is on its API now at $1.36 per million input tokens. Its benchmark scores are Mistral’s own and not yet checked independently.

**In short:** Mistral AI opened a public API preview of Mistral Large 4 on 6 October. It is a mixture-of-experts model with 1 trillion total parameters. Mistral says open weights will follow by the end of October.

## What launched

Mistral AI released Mistral Large 4 as a public preview on its API on 6 October 2026.[^1] Mistral describes it as a mixture-of-experts model with 1 trillion total parameters and 52 billion active parameters.[^1] Some coverage cites 49 billion active parameters from launch material, but Mistral's own page says 52 billion.[^1][^2]

Mistral says the model was trained from scratch on 3,800 NVIDIA Grace Blackwell GPUs in its own European data centres, and that it is fluent in more than 160 languages.[^1]

## Price

Mistral lists the preview at $1.36 per million input tokens and $4.18 per million output tokens.[^1] eWeek reports a two-week launch discount to $0.68 and $2.09, and notes that the list prices are well above Mistral Large 3's $0.50 and $1.50.[^2]

## Open weights

Mistral says downloadable weights are due by the end of October.[^1] VentureBeat reports a target date of 27 October and an expected custom Mistral licence.[^3] Mistral's announcement does not state the licence.

## The benchmark claims

Mistral reports 61.7% on DeepSWE v1.1 and 49.8% on a Coding Agent Index, among other scores.[^2] These are the company's own numbers. VentureBeat notes that several other models score higher on the live DeepSWE leaderboard, and that Large 4 was not yet listed on independent leaderboards when it reported.[^3]

## What it means for developers

The preview is available through Mistral's developer console. Teams already on Mistral Large 3 should compare the higher list price against their workloads before switching, and watch for the open-weights licence if they want to self-host.

## Fact ledger

- [Confirmed] Public API preview opened on 6 October 2026. (source 1)
- [Confirmed] Mixture-of-experts model with 1 trillion total and 52 billion active parameters. (source 1)
- [Confirmed] List price is $1.36 per 1M input tokens and $4.18 per 1M output tokens. (source 1)
- [Confirmed] Open weights are due by the end of October, Mistral says. (source 1)
- [Company claim] Trained on 3,800 NVIDIA Grace Blackwell GPUs in Europe. (source 1)
- [Company claim] Scores 61.7% on DeepSWE v1.1. (source 2)
- [Unconfirmed] A two-week launch discount brings prices to $0.68 and $2.09. (source 2)
- [Not known] The licence for the open weights.

## Sources

1. [Mistral AI: Introducing Mistral Large 4](https://mistral.ai/news/mistral-large-4/) — official
2. [eWeek: Mistral Large 4 enters preview](https://www.eweek.com/news/mistral-large-4-public-preview-enterprise-ai/)
3. [VentureBeat: Mistral debuts Large 4](https://venturebeat.com/technology/mistral-debuts-large-4-le-chonk-a-1-trillion-parameter-text-output-model-with-high-benchmarks-planned-for-open-weights-release)