LLM2014 Logic 2025-09: leaderboard
LLM2014 monthly logic benchmark for 2025-09, reporting aggregate logical-reasoning performance on the source score scale.
Metric: Median Score. Source: raw.githubusercontent.com. Saturation forecast: Around January 2027. 46 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | GPT-5 (High) | 84.1 | #98 (GPT-5) |
| 2 | GPT-5 Mini (High) | 78.12 | #186 (GPT-5 Mini) |
| 3 | Grok 4 | 74.79 | #181 |
| 4 | Grok 4 Fast (Reasoning) | 67.59 | #254 (Grok 4 Fast) |
| 5 | DeepSeek V3.1 (Thinking) | 54.13 | #264 (DeepSeek V3.1) |
| 6 | GPT-OSS-120B (High) | 53.51 | #341 (GPT-OSS-120B) |
| 7 | Qwen 3 235B A22B 2507 (Thinking) | 53.11 | #260 (Qwen 3 235B A22B 2507) |
| 8 | Gemini 2.5 Pro | 51.65 | #152 |
| 9 | Seed-1.6 (Thinking) | 49.28 | #265 (Seed-1.6) |
| 10 | Claude Sonnet 4 (Thinking) | 48.04 | #199 (Claude Sonnet 4) |
| 11 | Qwen 3 30B A3B 2507 (Thinking) | 46.52 | #376 (Qwen 3 30B A3B 2507) |
| 12 | Qwen 3 Next 80B A3B (Thinking) | 43.59 | #320 (Qwen 3 Next 80B A3B) |
| 13 | Qwen 3 235B A22B 2507 | 43.08 | #260 |
| 14 | Qwen 3 Max (Preview) | 42.33 | #230 |
| 15 | GLM 4.5 Air (Thinking) | 42.18 | #392 (GLM 4.5 Air) |
No result here: #3 Claude Opus 5.5, #6 GPT-6 Astra, #7 GPT-6.1 Sol.
Interactive version: theaggregate.ai/benchmark?slug=llm2014-logic-2025-09 · How It Works · Data refreshed daily, snapshot 2026-10-20.