HardcoreLogic - Increased Complexity: leaderboard
The Increased Complexity slice of HardcoreLogic — logic puzzles made harder by adding constraints and interactions; solve rate.
Metric: Accuracy (%). Source: huggingface.co. Saturation forecast: Around January 2027. 19 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | GPT-5 | 69.89 | #98 |
| 2 | Grok 4 | 58.26 | #181 |
| 3 | GPT-5 Mini | 55.76 | #186 |
| 4 | O4 Mini | 55.11 | #180 |
| 5 | GPT-OSS-120B | 54.08 | #341 |
| 6 | Grok 3 Mini | 48.48 | #281 |
| 7 | Qwen 3 235B A22B 2507 (Thinking) | 46.93 | #260 (Qwen 3 235B A22B 2507) |
| 8 | DeepSeek R1 0528 | 45.87 | #222 |
| 9 | DeepSeek V3.1 | 44.61 | #264 |
| 10 | Gemini 2.5 Pro | 43.8 | #152 |
| 11 | Qwen 3 Next 80B A3B (Thinking) | 41.97 | #320 (Qwen 3 Next 80B A3B) |
| 12 | Seed-OSS-36B-Instruct | 41.01 | #366 |
| 13 | Claude Sonnet 4 (Thinking) | 34.67 | #199 (Claude Sonnet 4) |
| 14 | Qwen 3 32B | 25.38 | #443 |
| 15 | Gemini 2.5 Flash | 25.11 | #245 |
No result here: #3 Claude Opus 5.5, #6 GPT-6 Astra, #7 GPT-6.1 Sol.
Interactive version: theaggregate.ai/benchmark?slug=hardcorelogic-increased-complexity · How It Works · Data refreshed daily, snapshot 2026-10-20.