Claude 3 Haiku — benchmark results

Provider: Anthropic. Released 2024-03-13. Access: API.

Unified ELO 1372 ± 8, rank #1333 of 1776 rated models, from 486 benchmark results.

Strongest benchmark results

BenchmarkScoreMetricPercentile
EIBench63.24Recall100
PECC27.67Pass@3100
OpenVLM MMT-Bench - Behavior Anomaly Detection75Score (%)97.8
OpenVLM MMT-Bench - Color Recognition90Score (%)97.8
OpenVLM MMT-Bench - Texture Material Recognition85Score (%)95.4
OpenVLM MMT-Bench - CIM73.3Score (%)94.7
OpenVLM MMT-Bench - Point Tracking85Score (%)93.9
OpenVLM MMT-Bench - Astronomical Recognition88.9Score (%)93
OpenVLM MMT-Bench - Color Assimilation75Score (%)93
OpenVLM MMT-Bench - Order Hallucination55Score (%)92.2
OpenVLM MMT-Bench - Polygon Localization65Score (%)91.7
EvoEval Creative47Pass@1 (%)91

Interactive version: theaggregate.ai/model?slug=claude-3-haiku · How the rankings work · Data refreshed daily, snapshot 2026-07-22.