VPCT

Predicting where a ball lands in a physics diagram.

Category	Multimodal
Metric	Accuracy
Source	Epoch AI (evaluator)
License	CC BY 4.0
Benchmark page	cbrower.dev
Status	Scored
Weight	0.6 within Multimodal
Fit	Discrimination 0.72, noise 0.31

Results

#	Model	Provider	Result	Date
1	Gemini 3 Pro Preview	Google	91.0%	2026-11-11
2	GPT-5.2 Extra High	OpenAI	84.0%	2026-11-11
3	Gemini 3 Flash Preview	Google	72.6%	2026-11-11
4	GPT-5.2 High	OpenAI	67.0%	2026-11-11
5	GPT-5 High	OpenAI	66.0%	2026-11-11
6	GPT-5 Medium	OpenAI	63.2%	2026-11-11
7	GPT-5.1 High	OpenAI	58.7%	2026-11-11
8	o4-mini Medium	OpenAI	57.5%	2026-11-11
9	GPT-5.1 Medium	OpenAI	53.3%	2026-11-11
10	o3 Medium	OpenAI	52.0%	2026-11-11
11	Gemini 2.5 Pro Preview 06-05	Google	48.0%	2026-11-11
12	Gemini 2.5 Flash (Sep 2025)	Google	46.2%	2026-11-11
13	GPT-4.5	OpenAI	45.0%	2026-11-11
14	gemini-robotics-er-1.5-preview	Unknown	40.8%	2026-11-11
15	GPT-5 Mini Medium	OpenAI	40.2%	2026-11-11
16	GPT-4o (2024-11-20)	OpenAI	40.0%	2026-11-11
17	Claude Opus 4.5 Thinking	Anthropic	40.0%	2026-11-11
18	Claude Sonnet 4.5 Thinking	Anthropic	39.8%	2026-11-11
19	Claude Sonnet 3.7	Anthropic	39.0%	2026-11-11
20	GPT-5 Mini High	OpenAI	39.0%	2026-11-11
21	Claude Sonnet 4.5	Anthropic	38.0%	2026-11-11
22	Gemini 2.5 Flash (Apr 2025)	Google	38.0%	2026-11-11
23	Claude Opus 4 Thinking	Anthropic	38.0%	2026-11-11
24	GPT-5 Nano High	OpenAI	37.2%	2026-11-11
25	o1 Medium	OpenAI	37.0%	2026-11-11
26	GPT-5 Nano Medium	OpenAI	35.4%	2026-11-11
27	Claude Opus 4.1	Anthropic	35.0%	2026-11-11
28	Claude Sonnet 3.7 Thinking	Anthropic	35.0%	2026-11-11
29	GPT-4o mini	OpenAI	34.0%	2026-11-11
30	Claude Sonnet 4 Thinking	Anthropic	34.0%	2026-11-11
31	Claude Opus 4	Anthropic	33.0%	2026-11-11
32	Claude Sonnet 3.5 v2	Anthropic	33.0%	2026-11-11
33	Claude 3.5 Sonnet	Anthropic	33.0%	2026-11-11
34	Claude Opus 4.1 Thinking	Anthropic	33.0%	2026-11-11
35	Claude Sonnet 4	Anthropic	30.0%	2026-11-11
36	Gemini 2.5 Flash-Lite Preview 09-2025	Google	30.0%	2026-11-11