DiningBench - Calorie Estimation: leaderboard

DiningBench (2026) evaluates vision-language models on multi-view user photos of restaurant dishes, testing dish classification against same-menu distractors, nutrition estimation and food visual question answering. This board is the mean absolute percentage error of the estimated calories; lower is better.

Metric: Mean absolute percentage error (%) of the estimated calories over the 1,650 DiningBench nutrition-estimation dishes (multi-view restaurant photos; ground truth from merchant labels or ingredient-based estimates cross-checked against USDA FoodData Central); lower is better. Source: arxiv.org. Saturation forecast: Around March 2027. 29 models tracked. Updated 7 Oct 2026.

Top models

#ModelScoreOverall rank
1Gemini 3 Pro (Preview)19.99#123
2Gemini 3 Flash (Preview)20.57#156
3GPT-525.14#185
4O4 Mini27.68#372
5Gemma 3 12B (IT)27.95#1907
6Gemini 2.5 Pro28.61#280
7Gemini 2.5 Flash30.21#466
8Qwen 3 VL 8B Instruct31.06#1043
9Qwen 3 VL 30B A3B (Thinking)32.33#727
10Qwen 3 VL 30B A3B Instruct32.49#903
11Qwen 2.5 VL 32B Instruct32.66#1065
12MiniCPM-V-4.534.79#1451
13Claude Sonnet 4.535.01#267
14Qwen 2.5 VL 72B Instruct35.88#849
15GPT-4.136.88#506

No result here: #2 Claude Fable 5.2, #3 Gemini 4 Neon, #4 GPT-6.2 Astra.

Interactive version: theaggregate.ai/benchmark?slug=diningbench-calorie-estimation · How It Works · Data refreshed daily, snapshot 2026-11-22.