AGI-Eval Community Base - Learning: leaderboard

AGI-Eval Community private-set Base Language Models evaluation (Learning) measuring accuracy (%) across bilingual capability dimensions.

Metric: Accuracy (%). Source: agi-eval.org. Saturation forecast: Estimated already saturated. 25 models tracked. Updated 22 Nov 2026.

Top models

#ModelScoreOverall rank
1DeepSeek-V3.1-Base86.91
2DeepSeek V3 Base86.33
3Kimi K2 Base86.17#939
4GLM-4.5-Base85.29
5DeepSeek-V3.2-Exp-Base84.28
6Llama 3.1 405B Base80.07
7Qwen 2.5 32B (Base)76.73
8Llama 4 Maverick Base76.42
9ERNIE-4.5-300B-A47B-Base75.96
10Qwen3-4B-Base74.46#2439
11Qwen3-8B-Base74.44#1754
12Qwen3-14B-Base74.39#1473
13Llama 3.1 70B Base72.86
14Qwen3-30B-A3B-Base71.91#1475
15Qwen 2.5 7B (Base)71.09

No result here: #2 Claude Fable 5.2, #3 Gemini 4 Neon, #4 GPT-6.2 Astra.

Interactive version: theaggregate.ai/benchmark?slug=agi-eval-community-base-learning · How It Works · Data refreshed daily, snapshot 2026-11-22.