Yi Large: benchmark results

01.AI's closed API flagship chat model with strong multilingual performance, the Chinese startup's top proprietary offering of 2024. Provider: 01.AI. Released 2024-06-25. Access: API.

Unified ELO 1516 ± 1, rank #625 of 1392 rated models, from 15 benchmark results.

Strongest benchmark results

BenchmarkScoreMetricPercentile
BenchBench78.89Aggregate Score (%)83.8
MMLU79.3Accuracy (%)79.6
WildBench48.93WB Score Task-Macro79
BigCodeBench37.7Pass@1 (%)47.2
ZebraLogic18.8Puzzle Accuracy (%)42.6
BenchTable42.8Total Score (%)41
AI for Education Pedagogy - Science74.86Accuracy (%)35.7
AI for Education Pedagogy - Social studies70Accuracy (%)33.4
TextClass Benchmark1473.21Meta-Elo (self-reported)31.8
AI for Education Pedagogy71.52Accuracy (%)29.6
AI for Education Pedagogy - Secondary70.6Accuracy (%)29.4
AI for Education Pedagogy - Maths65.87Accuracy (%)25.8

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