HumorRank - Humor Transfer Bench: leaderboard
HumorRank (2026) ranks humor generation with a full round-robin tournament of pairwise joke comparisons judged by Llama 3.3 70B Instruct. This board is the Bradley-Terry rating on Humor Transfer Bench prompts, an open Elo-style scale whose ratings average 1000; higher is better.
Metric: Bradley-Terry rating (open scale, mean about 1000) from a full round-robin of pairwise joke comparisons on Humor Transfer Bench (HTB), 14,400 judgments, judged by Llama 3.3 70B Instruct; higher is better. Source: arxiv.org. Saturation forecast: Estimated already saturated. 9 models tracked. Updated 7 Oct 2026.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | GPT-5 | 1314.7 | #178 |
| 2 | Kimi K2 | 1242 | #443 |
| 3 | HumorGen-7B | 1097.7 | |
| 4 | Claude 3.5 Haiku | 1054.2 | #1580 |
| 5 | Gemini 2.5 Pro | 1024 | #276 |
| 6 | GPT-OSS-120B | 1009.6 | #704 |
| 7 | Qwen 3 32B | 942.5 | #972 |
| 8 | Llama 3.3 70B | 791.9 | #1514 |
| 9 | Qwen 2.5 7B Instruct | 523.5 | #2723 |
No result here: #1 Claude Fable 5.2, #3 Gemini 4 Neon, #4 GPT-6.2 Astra.
Interactive version: theaggregate.ai/benchmark?slug=humorrank-humor-transfer-bench · How It Works · Data refreshed daily, snapshot 2026-11-21.