Pixtral Large — benchmark results

Mistral's 124B frontier multimodal model (November 2024) built on Mistral Large 2, with weights released under the research-only MRL license. Provider: Mistral. Released 2024-11-18. Access: Open.

Unified ELO 1484 ± 33, rank #860 of 1776 rated models, from 38 benchmark results.

Strongest benchmark results

BenchmarkScoreMetricPercentile
LLM Stats (MM-MT-Bench)74Score (%)93.8
LLM Stats (AI2D)93.8Score (%)90.3
LLM Stats (ChartQA)88.1Score (%)73.9
LLM Stats (MathVista)69.4Score (%)62.2
LLM Stats (DocVQA)93.3Score (%)58
AI for Education Visual Reasoning - match (process)33.3Accuracy (%)46.8
AA TAU-2 Bench36.55Accuracy (%)44.4
AA SciCode29.17Accuracy (%)41.2
AA MMLU-Pro70.11Accuracy (%)36.6
ProLLM - Image Understanding55Score (%)34.6
ZEROBench-Sub18.68Score (self-reported)33.3
AA MATH-50071.4Accuracy (%)32.4

Interactive version: theaggregate.ai/model?slug=pixtral-large · How the rankings work · Data refreshed daily, snapshot 2026-07-22.