AutoBench: leaderboard
Automated multi-metric LLM evaluation ranking models by cost, latency, and domain performance. Models ranked by aggregate score across practical business tasks.
Metric: Judge Rating (1-5). Source: autobench.org. Saturation forecast: Around January 2027. 38 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | GPT-5.2 Pro | 4.48 | #61 |
| 2 | GPT-5.2 | 4.43 | #110 |
| 3 | Gemini 3 Pro (Preview) | 4.41 | #66 |
| 4 | Claude Opus 4.5 | 4.39 | #83 |
| 5 | GPT-5.1 | 4.38 | #135 |
| 6 | Kimi K2 (Thinking) | 4.32 | #243 (Kimi K2) |
| 7 | Claude Sonnet 4.5 | 4.3 | #142 |
| 8 | Gemini 3 Flash (Preview) | 4.3 | #82 |
| 9 | Gemini 2.5 Pro | 4.29 | #152 |
| 10 | GPT-5 Mini | 4.29 | #186 |
| 11 | Grok 4.1 Fast (Reasoning) | 4.21 | #217 (Grok 4.1 Fast) |
| 12 | Grok 4 | 4.2 | #181 |
| 13 | Qwen 3 235B A22B 2507 (Thinking) | 4.2 | #260 (Qwen 3 235B A22B 2507) |
| 14 | GPT-OSS-120B | 4.18 | #341 |
| 15 | Gemini 2.5 Flash | 4.17 | #245 |
No result here: #3 Claude Opus 5.5, #6 GPT-6 Astra, #7 GPT-6.1 Sol.
Interactive version: theaggregate.ai/benchmark?slug=autobench · How It Works · Data refreshed daily, snapshot 2026-10-20.