OST-Bench - Agent Object Spatial: leaderboard

The Agent Object Spatial questions of OST-Bench's multi-round version: the model judges the direction and distance of previously seen objects relative to its current position. Scores are accuracy (%); the README states 1.4k scenes and 10k question-answer pairs, the per-model sample is not stated.

Metric: Accuracy (%). Source: github.com. Saturation forecast: Estimated already saturated. 10 models tracked. Updated 17 Nov 2026.

Top models

#ModelScoreOverall rank
1GPT-4.137.65#505
2LLaVA-Video-72B-Qwen235.07#1344
3InternVL2.5-8B34.61#2401
4InternVL2.5-38B33.95#1366
5Gemini 2.0 Flash33.8#723
6InternVL2.5-78B32.93#1036
7Qwen 2.5 VL 32B Instruct32.18#1074
8Qwen 2.5 VL 7B Instruct31.53#1893
9Qwen 2.5 VL 72B Instruct28.87#849
10LLaVA-Video-7B-Qwen228.8#1910

No result here: #1 Claude Fable 5.2, #2 GPT-6.2 Astra, #3 Gemini 4 Carbon.

Interactive version: theaggregate.ai/benchmark?slug=ost-bench-agent-object-spatial · How It Works · Data refreshed daily, snapshot 2026-11-17.