CommunityBench - Community Identification: leaderboard
Metric: Accuracy (%). Source: arxiv.org. Saturation forecast: Around December 2026. 17 models tracked.
Top models
| # | Model | Score |
|---|---|---|
| 1 | GPT-4o | 84.3 |
| 2 | DeepSeek V3 (0324) | 82.32 |
| 3 | Grok 4 | 82.23 |
| 4 | Llama 3.1 70B Instruct | 77.02 |
| 5 | Llama 3.3 70B Instruct | 73.8 |
| 6 | Qwen 2.5 72B Instruct | 68.96 |
| 7 | Qwen 2.5 14B Instruct | 68.53 |
| 8 | GLM-4 32B (0414) | 67.18 |
| 9 | Qwen 3 32B | 60.86 |
| 10 | Qwen 2.5 7B Instruct | 58.66 |
| 11 | GLM-4 9B (0414) | 51.25 |
| 12 | InternLM3-8B-Instruct | 49.24 |
| 13 | Mistral 7B Instruct (v0.3) | 48.55 |
| 14 | Qwen 3 14B | 48.52 |
| 15 | Llama 3.1 8B Instruct | 47.37 |
Interactive version: theaggregate.ai/benchmark?slug=communitybench-community-identification · How It Works · Data refreshed daily, snapshot 2026-09-25.