LLM-WikiRace: leaderboard
Planning benchmark where models navigate Wikipedia hyperlink graphs from a start page to a target page, testing strategy over real-world knowledge graphs.
Metric: Average Success (%). Source: llmwikirace.github.io. Saturation forecast: Around May 2027. 30 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | GPT-5 | 55.83 | #98 |
| 2 | Claude Opus 4.5 | 55.17 | #83 |
| 3 | Claude Opus 4.6 | 54.57 | #63 |
| 4 | DeepSeek R1 | 54.23 | #251 |
| 5 | Gemini 2.5 Flash | 51.4 | #245 |
| 6 | GPT-5.2 | 50.23 | #110 |
| 7 | GPT-5 Mini | 47.5 | #186 |
| 8 | Claude Sonnet 4.5 | 47.3 | #142 |
| 9 | Kimi K2 | 46.6 | #243 |
| 10 | Grok 4.1 Fast | 46.57 | #217 |
| 11 | Gemini 2.0 Flash | 45.1 | #342 |
| 12 | Llama 3 70B | 43.6 | #631 |
| 13 | Gemma 3 27B (IT) | 36.67 | #523 |
| 14 | GPT-5 Nano | 33.4 | #425 |
| 15 | Gemma 3 12B (IT) | 32.07 | #664 |
No result here: #3 Claude Opus 5.5, #6 GPT-6 Astra, #7 GPT-6.1 Sol.
Interactive version: theaggregate.ai/benchmark?slug=llm-wikirace · How It Works · Data refreshed daily, snapshot 2026-10-20.