BARISTA - Temporal VQA: leaderboard

BARISTA (2026) is an egocentric benchmark of 185 coffee-preparation videos with verified per-frame scene graphs, from which zero-shot language tasks for vision-language models are derived. This board is the mean of answer accuracy and completeness on questions about short clips, judged by Gemini 2.5 Flash on a 0 to 1 scale; higher is better.

Metric: Mean of answer accuracy and completeness, Gemini 2.5 Flash G-Eval judge (0 to 10 rubric rescaled to 0-1), on 300 six-frame clips with reference answers generated once by Gemini 2.5 Flash Lite, on BARISTA, an egocentric coffee-preparation video dataset with verified per-frame scene graphs, zero-shot with shared task prompts, temperature 0.7 (GPT-5.4 mini at its fixed sampling), low thinking effort for the closed models, mean of three runs; higher is better. Source: arxiv.org. Saturation forecast: Around February 2027. 5 models tracked. Updated 7 Oct 2026.

Top models

#ModelSettingScoreOverall rank
1Qwen 3.5 27B0.701#375
2GPT-5.4 MiniLow0.686#406
3Gemini 3 FlashLow0.665#175
4Qwen 3.5 9B0.655#828
5Gemma 4 31B0.649#382

No result here: #1 Claude Fable 5.2, #3 Gemini 4 Neon, #4 GPT-6.2 Astra.

Interactive version: theaggregate.ai/benchmark?slug=barista-temporal-vqa · How It Works · Data refreshed daily, snapshot 2026-11-21.