LOL Arena: leaderboard
Select the funniest response in Cards Against Humanity-style rounds, scored by agreement with the human judge. Current Inspect pipeline, lowest reasoning level offered by each API; 14,459 nominal human-judge rounds.
Metric: Human judge-choice agreement (%), current Inspect pipeline; lowest API reasoning. Source: goodstartlabs.com. Saturation forecast: Around 2034. 36 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | GPT-6 Astra | 50.3 | #6 |
| 2 | Claude Fable 5.1 | 50 | #9 |
| 3 | Gemini 3.7 Flash | 49.5 | #23 |
| 4 | Claude Fable 5 | 49.4 | #16 |
| 5 | Claude Opus 5.5 | 49.3 | #3 |
| 6 | Muse Spark 1.3 | 49.1 | #24 |
| 7 | Grok 4.6 | 49 | #26 |
| 8 | GPT-4.1 | 48.9 | #248 |
| 9 | Gemini 3.1 Pro (Preview) | 48.8 | #58 |
| 10 | GPT-5.6 Sol | 48.8 | #19 |
| 11 | Claude Opus 5 | 48.7 | #12 |
| 12 | Muse Spark 1.2 | 48.6 | #62 |
| 13 | GPT-5 | 48.5 | #98 |
| 14 | Claude Opus 4.8 | 48.4 | #37 |
| 15 | GPT-4o | 48.2 | #343 |
No result here: #7 GPT-6.1 Sol, #8 Gemini 4 Argon, #10 Muse Wonder Full.
Interactive version: theaggregate.ai/benchmark?slug=lol-arena · How It Works · Data refreshed daily, snapshot 2026-10-20.