// BENCHMARK

AGIEval benchmark

AI model leaderboard for the AGIEval benchmark. Compare how large language models score on AGIEval, see the full ranking, and understand what this AI benchmark measures. Qwen3-235B-A22B-Thinking (Jul 2025) currently leads with 90.01. Human-centric exam benchmark — Chinese gaokao, SAT, LSAT, lawyer-qualification etc.

Leaderboard

# Model Organization Score Variant Source
#1 Qwen3-235B-A22B-Thinking (Jul 2025) Qwen 90.01 thinking-cited-ring-1t official ↗
#2 DeepSeek-V3.1-Terminus DeepSeek 89.83 thinking-cited-ring-1t official ↗
#3 gemini-2.5-pro Google DeepMind 88.99 cited-ring-1t official ↗
#4 GPT-5 OpenAI 88.56 thinking-high-cited-ring-1t official ↗
#5 Ring-1T Ant Group 88.13 thinking official ↗
#6 Granite-4.0-H-Small IBM 70.63 0-shot-cot official ↗
#7 OLMo 2 32B Allen Institute for AI 67 base-from-hf-readme official ↗
#8 Qwen2.5-7B Qwen 63.7 base-cited-olmo2 official ↗
#9 Granite-4.0-H-Tiny IBM 62.15 0-shot-cot official ↗
#10 Granite-4.0-H-Micro IBM 59 0-shot-cot official ↗
#11 Gemma 2 9B Google DeepMind 57.3 base-cited-olmo2 official ↗
#12 Llama 3.1-8B Meta AI 51.3 base-cited-olmo2 official ↗
#13 Llama 2-13B Meta AI 41.5 base-cited-olmo2 official ↗

Frequently asked questions about AGIEval

What is the AGIEval benchmark?

Human-centric exam benchmark — Chinese gaokao, SAT, LSAT, lawyer-qualification etc.

How is the AGIEval benchmark scored?

AGIEval is scored using the accuracy metric, where a higher score is better. GenAIList aggregates reported scores from model providers and papers into a single ranked leaderboard.

Which AI model scores highest on AGIEval?

As of the latest reported scores on GenAIList, Qwen3-235B-A22B-Thinking (Jul 2025) achieves the highest result on AGIEval with a score of 90.01.

Is a higher AGIEval score better?

Yes. On AGIEval a higher score indicates better performance, so models near the top of the leaderboard are the strongest.