AGI-Eval Community - Instruction Following: leaderboard
AGI-Eval community leaderboard (agi-eval.cn), the private-set LLM board built from the platform's non-public items; latest monthly snapshot. Unrelated to the AGIEval exam dataset. This board is the instruction following dimension; upstream's Chinese split is identical, so only this column is kept. Upstream accuracy 0-1, shown as 0-100; higher is better.
Metric: Accuracy (%). Source: agi-eval.cn. Saturation forecast: Around February 2027. 141 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | GPT-5.5 | 71.29 | #33 |
| 2 | GPT-5.4 | 69.47 | #83 |
| 3 | GLM-5.2 | 69.23 | #80 |
| 4 | Qwen 3.7 Max | 68.99 | #85 |
| 5 | Gemini 3.5 Flash | 67.09 | #66 |
| 6 | Seed 2.1 Pro | 67.09 | #73 |
| 7 | Kimi K2.6 | 65.74 | #113 |
| 8 | Hy3 | 65.58 | #100 |
| 9 | O3 Pro | 64.63 | #116 |
| 10 | Kimi K3 | 64.23 | #35 |
| 11 | GPT-5.1 | 64.08 | #145 |
| 12 | DeepSeek V4 Pro (Max) | 64.08 | #112 (DeepSeek V4 Pro) |
| 13 | Gemini 3.1 Pro (Preview) | 64 | #64 |
| 14 | Claude Opus 4.8 | 63.52 | #39 |
| 15 | Seed 2.0 Pro | 63.52 | #122 |
No result here: #3 Claude Opus 5.5, #6 GPT-6 Astra, #7 GPT-6.1 Sol.
Interactive version: theaggregate.ai/benchmark?slug=agi-eval-community-instruction-following · How It Works · Data refreshed daily, snapshot 2026-11-05.