Open weightsMistral AI/10 trustVerified

Mistral Large 2

Name: Mistral Large 2
Rating: 8

Mistral's 123B flagship that punched at GPT-4o level from a single node

About

Mistral Large 2 arrived July 24, 2024 as a 123B dense transformer designed to run inference on a single node. At launch Mistral positioned it alongside GPT-4o, Claude 3 Opus, and Llama 3 405B on capability, with particular strength in code (80+ languages) and multilingual reasoning across dozens of languages. A 24.11 refresh in November 2024 improved it further. The weights are downloadable but under the Mistral Research License, which restricts commercial use, so most production usage goes through the Mistral API or cloud partners at $2/$6 per million tokens. It has aged well as a European alternative to the US hyperscaler flagships, though newer reasoning models have moved the frontier past it.

Specs

Context window

128,000 tokens

Pricing

$2.00 per 1M input tokens, $6.00 per 1M output tokens

Free quota

Limited free tier on Mistral's La Plateforme; API is pay-as-you-go

API access

Mistral API (La Plateforme), Azure AI, AWS Bedrock, Google Vertex AI; weights downloadable under Mistral Research License

Weights

Open weights

Released

July 24, 2024

Benchmarks

Reported: MMLU 84.0% (5-shot); team-positioned on par with GPT-4o and Claude 3 Opus at launch

Provider-reported unless a source says otherwise. Benchmarks are marketing-adjacent by nature — treat them as directional.

Use cases

Multilingual chat and contentCode generation across 80+ languagesEU-hosted deploymentsSingle-node self-hosted inference

Trust score — 8/10

/10 trustVerified

Scored 8/10 (verified tier, metrics-v1). Evidence: July 24, 2024 release, 123B dense, 128K context, $2/$6 pricing corroborated across 3 sources; Mistral Research License (non-commercial) captured honestly; openWeights set true with caveat; 24.11 refresh noted; MMLU 84.0% labeled as reported; launch positioning vs GPT-4o/Opus attributed to Mistral.