Tencent HY 2.0 Think: benchmark results

Provider: Other. Released 2025-12-05. Access: API.

Unified ELO 1780 ± 37, rank #182 of 2656 rated models, from 40 benchmark results.

Strongest benchmark results

BenchmarkScoreMetricPercentile
OpenCompass Math - College (CompassBench 2601)75.9Score (%)92
OpenCompass Agent - Tool Use (CompassBench 2601)65.8Score (%)84
OpenCompass Knowledge - Humanities (CompassBench 2601)93.5Score (%)84
OpenCompass Knowledge - Common Sense (CompassBench 2601)93Score (%)80
OpenCompass Knowledge - Science (CompassBench 2601)94.2Score (%)72
OpenCompass LLM - Agent (CompassBench 2601)68.9Score (%)72
OpenCompass LLM - Math (CompassBench 2601)69.4Score (%)72
OpenCompass LLM - Knowledge (CompassBench 2601)91.8Score (%)68
OpenCompass LLM - Average (CompassBench 2601)69.1Score (%)64
OpenCompass Math - Competition (CompassBench 2601)62.9Score (%)60
OpenCompass LLM - Reasoning (CompassBench 2601)48.3Score (%)56
OpenCompass Reasoning - Common (CompassBench 2601)59.6Score (%)56

Interactive version: theaggregate.ai/model?slug=tencent-hy-2-0-think · How It Works · Data refreshed daily, snapshot 2026-09-19.