LithoBench - Comprehensive Reasoning (Open-Ended): leaderboard
Metric: OPEN_J score (0-10) = 0.4 x ROUGE-L x 10 + 0.2 x BLEU-4 x 10 + 0.4 x the GPT-5.4 judge's overall score (0-10), comparing the answer with the reference answer, on the 1,200 open-ended comprehensive reasoning questions of LithoBench (multi-step geological reasoning in GF-2 remote-sensing image patches); temperature 0, 1,024 output tokens; higher is better. Source: arxiv.org. Saturation forecast: Around 2030. 12 models tracked.
Top models
| # | Model | Score |
|---|---|---|
| 1 | Qwen 3.5 397B A17B | 4.64 |
| 2 | Kimi K2.5 | 4.55 |
| 3 | GLM-4.6V | 4.45 |
| 4 | Grok 4 | 4.41 |
| 5 | Gemini 3 Pro | 4.17 |
| 6 | Qwen 3.5 9B | 3.62 |
| 7 | Claude Sonnet 4.6 | 3.16 |
| 8 | Qwen 3.5 27B | 2.04 |
Interactive version: theaggregate.ai/benchmark?slug=lithobench-comprehensive-reasoning-open-ended · How It Works · Data refreshed daily, snapshot 2026-10-07.