FlavourBench - Pairing: leaderboard
FlavourBench pairing family: 178 of the 534 complete-common-core culinary tasks, scored by executable reward maps (arXiv 2608.20574v3). Family score 0-100, higher is better, from family_scores in data-complete-core/leaderboard.jsonl in the josefchen/flavourbench dataset; each route's recorded reasoning effort is kept in the model label.
Metric: Pairing Family Score (0-100). Source: huggingface.co. Saturation forecast: Around April 2028. 27 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Kimi K3 (Low) | 73.72 | #35 (Kimi K3) |
| 2 | MiniMax-M3 | 69.91 | #151 |
| 3 | GLM-5.3 | 69.59 | #44 |
| 4 | GLM-5.2 (Non-reasoning) | 67.24 | #80 (GLM-5.2) |
| 5 | Gemini 3.6 Flash (Minimal) | 66.22 | #57 (Gemini 3.6 Flash) |
| 6 | Mistral Large 3 | 65.84 | #410 |
| 7 | Nemotron 3.5 Lightning | 64.33 | #461 |
| 8 | Command A | 63.86 | #498 |
| 9 | Llama 4 Maverick | 61.39 | #486 |
| 10 | Command-R+ (08-2024) | 54.86 | #838 |
No result here: #3 Claude Opus 5.5, #6 GPT-6 Astra, #7 GPT-6.1 Sol.
Interactive version: theaggregate.ai/benchmark?slug=flavourbench-pairing · How It Works · Data refreshed daily, snapshot 2026-11-05.