EgoPoint-Bench: leaderboard

EgoPoint-Bench (2026) asks multimodal models about the object a first-person hand points at, on held-out simulation scenes and real smart-glasses recordings, zero-shot; true/false and single-choice answers by exact match and open answers scored by a GPT-4o judge. This board is the item-weighted accuracy over both test sets, in percent; higher is better.

Metric: Accuracy (%) over all 3,091 test questions (simulation and real-world, item-weighted) about the object a first-person hand points at; zero-shot (direct inference); true/false and single-choice questions by exact match, open-ended answers scored 0 to 1 by a GPT-4o judge; greedy decoding for open models, temperature 0 for API models; higher is better. Source: arxiv.org. Saturation forecast: Around 2029. 12 models tracked. Updated 7 Oct 2026.

Top models

#ModelScoreOverall rank
1Gemini 3 Flash62.71#180
2Gemini 3 Pro62.29#146
3Qwen 3 VL 32B Instruct60.54#607
4GPT-5.2 Instant59.29#309
5GPT-5 Mini58.75#373
6Qwen 3 VL 8B Instruct58.23#1043
7GLM-4.6V-Flash54.47#1049
8InternVL3.5-8B-Instruct54.3#1911
9InternVL3.5-14B-Instruct53.23#1747
10InternVL3.5-2B-Instruct52.49#6576
11LLaVA-v1.5-7B48.21#7642
12LLaVA-NeXT-7B47.52#5045

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

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