ELBench - General Capability: leaderboard
The general-capability module of ELBench (Jiang et al., 2026): 894 items sampled from MMLU-Pro, C-Eval, IFEval, MATH-500 and AIME 2024 to 2026, scored by reference matching or deterministic rules. Mean per-item score in percent, higher is better.
Metric: Module score (%; mean per-item score on 894 items sampled from MMLU-Pro, C-Eval, IFEval, MATH-500 and AIME 2024-2026, scored by reference matching or deterministic rules; zero-shot, temperature 0; open-ended items scored by a Qwen3.6 rubric judge, majority of 9 calls). Source: arxiv.org. Saturation forecast: Estimated already saturated. 9 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Gemini 3.5 Flash | 93.4 | #101 |
| 2 | Claude Opus 4.8 | 91.9 | #56 |
| 3 | DeepSeek V4 Flash | 88.5 | #258 |
| 4 | DeepSeek V4 Pro | 88.2 | #188 |
| 5 | Seed 2.0 Pro | 86.6 | #211 |
| 6 | GPT-5.4 | 86.4 | #123 |
| 7 | GLM-5.1 | 83.2 | #179 |
No result here: #1 Claude Fable 5.2, #2 Gemini 4 Carbon, #5 Claude Opus 5.5.
Interactive version: theaggregate.ai/benchmark?slug=elbench-general-capability · How It Works · Data refreshed daily, snapshot 2026-11-13.