Llama 4 Maverick Instruct: benchmark results

Meta's natively multimodal open MoE (400B total/17B active, 128 experts) with a 1M-token context (April 2025). Provider: Meta. Released 2025-04-05. Access: Open.

Unified ELO 1564 ± 1, rank #386 of 1392 rated models, from 125 benchmark results.

Strongest benchmark results

BenchmarkScoreMetricPercentile
SEA LLM Leaderboard - SeaExam76.7Private Average Score (%)92.2
KOFFVQA - Relationship74.33Score (%)85.2
KOFFVQA - Graph and Chart Understanding83.33Score (%)84.6
KOFFVQA - Hallucination and Robustness80Score (%)80.9
KOFFVQA - Object Attributes79.17Score (%)79
KOFFVQA - Document Understanding86Score (%)78.4
MERA v2 - Enantiosemy49.6Score (%)77.8
Vals AI MGSM92.44Accuracy (%)77.3
Enkrypt AI - Jailbreak Risk4.54Risk Score76.9
UGI - Natural Intelligence30.93NatInt Score76
ProLLM - LLM-as-a-Judge80.9Score (%)75.9
KOFFVQA76.95Overall Score (%)75.3

Interactive version: theaggregate.ai/model?slug=llama-4-maverick-instruct · How It Works · Data refreshed daily, snapshot 2026-09-05.