Subquadratic LLM Leaderboard — leaderboard
Benchmark for non-transformer architectures (RWKV, Mamba, SSMs): evaluates subquadratic-attention models across ARC, HellaSwag, MMLU, TruthfulQA, WinoGrande, and GSM8K.
Metric: Average Score. Source: huggingface.co. 39 models tracked.
Top models
| # | Model | Score |
|---|---|---|
| 1 | v5-EagleX-v2-7B-HF | 48.66 |
| 2 | EagleX_1-7T_Chat | 47.36 |
| 3 | EagleX_1-7T | 46.84 |
| 4 | Hermes-RWKV-v5-7B | 46.42 |
| 5 | rwkv-5-world-7b | 45.15 |
| 6 | bagel-dpo-2.8b-v0.2 | 43.89 |
| 7 | bagel-2.8b-v0.2 | 43.47 |
| 8 | rwkv-raven-14b | 42.09 |
| 9 | Hermes-mamba-2.8b-slimpj-cDPO | 41.77 |
| 10 | Hermes-mamba-2.8b-slimpj | 41.65 |
Interactive version: theaggregate.ai/benchmark?slug=subquadratic-llm-leaderboard · How the rankings work · Data refreshed daily, snapshot 2026-07-22.