RideWay - Success Rate: leaderboard
Metric: Success rate (%; share of trials passing the three-judge LLM success gate over the pooled 58-task suite, 3 attempts per model). Source: arxiv.org. Saturation forecast: Around December 2026. 24 models tracked.
Top models
| # | Model | Score |
|---|---|---|
| 1 | MiMo-V2.5-Pro (Non-reasoning) | 87.4 |
| 2 | Claude Sonnet 5 | 82.8 |
| 3 | Gemini 3.5 Flash (Minimal) | 82.2 |
| 4 | Claude Opus 4.6 | 81.6 |
| 5 | DeepSeek V4 Pro (Non-reasoning) | 81 |
| 6 | GLM-5.2 (Non-reasoning) | 78.7 |
| 7 | Qwen 3.6 27B (Non-reasoning) | 77 |
| 8 | Gemma 4 31B | 75.9 |
| 9 | GPT-5.5 (Non-reasoning) | 75.3 |
| 10 | Qwen 3.7 Max (Non-reasoning) | 74.1 |
| 11 | Qwen 3.6 Plus (Non-reasoning) | 71.8 |
| 12 | Kimi K2.6 (Non-reasoning) | 61.5 |
| 13 | Gemma 4 26B | 60.3 |
| 14 | Qwen 3.5 9B (Non-reasoning) | 59.2 |
| 15 | Qwen 3.6 35B A3B (Non-reasoning) | 56.3 |
Interactive version: theaggregate.ai/benchmark?slug=rideway-success-rate · How It Works · Data refreshed daily, snapshot 2026-09-26.