LLMEval-Logic Formalization Fixed: leaderboard
LLMEval logic-reasoning leaderboard (llmeval.com) — formalization track with a fixed formal template; accuracy (%).
Metric: Accuracy (%). Source: llmeval.com. Saturation forecast: Around April 2028. 14 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Kimi K2.5 (Thinking) | 59.4 | #144 (Kimi K2.5) |
| 2 | Gemini 3.1 Pro (Preview) (Thinking) | 58.1 | #58 (Gemini 3.1 Pro (Preview)) |
| 3 | Claude Opus 4.6 (Thinking) | 56.9 | #63 (Claude Opus 4.6) |
| 4 | Seed 2.0 Pro (Thinking) | 56.9 | #118 (Seed 2.0 Pro) |
| 5 | Gemini 3.1 Pro (Preview) (Low) | 54.5 | #58 (Gemini 3.1 Pro (Preview)) |
| 6 | Qwen 3.5 Plus (Thinking) | 52.4 | #125 (Qwen 3.5 Plus) |
| 7 | Hy3-preview (Reasoning) | 51.2 | #157 (Hy3-preview) |
| 8 | Claude Opus 4.6 (Non-reasoning) | 50 | #63 (Claude Opus 4.6) |
| 9 | Seed 2.0 Pro (Non-reasoning) | 47.6 | #118 (Seed 2.0 Pro) |
| 10 | Kimi K2.5 (Non-reasoning) | 38.6 | #144 (Kimi K2.5) |
| 11 | Qwen 3.5 Plus (Non-reasoning) | 35.8 | #125 (Qwen 3.5 Plus) |
| 12 | Hy3-preview (Non-reasoning) | 24.8 | #157 (Hy3-preview) |
No result here: #3 Claude Opus 5.5, #6 GPT-6 Astra, #7 GPT-6.1 Sol.
Interactive version: theaggregate.ai/benchmark?slug=llmeval-logic-formalization-fixed · How It Works · Data refreshed daily, snapshot 2026-10-20.