Benchmark Almanac

VPCT

Gemini 3 Pro 91% leads since November 2025
Gemini 3 Pro 91% leads since November 2025
Fading

Models view an image of a ball, ramps, and three buckets and predict which bucket the ball falls into. Scored by accuracy on 100 problems.

Made by
Chase Brower
Released
January 2025
Kind
Physical prediction
Level
Everyday
Scored
Checked automatically
Answers
Multiple choice
Tasks
All public
Strongest model tested
GPT-5.2 #50 on the capability index
Last in a launch post
December 2025, dropped after it saturated
Last model added
March 2026
Scores refreshed
November 12th, 2026
27
Models scored on a board Epoch AI copied
1
Launch post lists it, from 1 lab
2
System cards report it, from 1 lab

Gemini 3 Pro leads with 91% among models released by December 2025

Best score per model from VPCT leaderboard, as Epoch AI copied it. We don’t read that board ourselves yet. Where a model ran with several settings, we keep its best.

Gemini 3 Pro leads with 91%, claims included

Best score per model from every source, including labs’ own numbers and claims Epoch AI collected from reports and papers. Labs pick their own settings, so a claim can beat an independent run of the same model.

The newest model with a score here came out in December 2025, and nobody has tested a newer one. Among the 20 highest-ranked models on the capability index, these have no score yet:

27 models. Scroll sideways for the rest.

  1. 91.0 Gemini 3 Pro
  2. 84.0 GPT-5.2
  3. 72.6 Gemini 3 Flash
  4. 66.0 GPT-5
  5. 58.7 GPT-5.1
  6. 57.5 o4-mini
  7. 52.0 o3
  8. 48.0 Gemini 2.5 Pro
  9. 46.4 Gemini 2.5 Pro
  10. 46.2 Gemini 2.5 Flash
  11. 45.0 GPT-4.5
  12. 40.8 gemini-robotics-er-1.5-preview
  13. 40.5 Gemini 2.5 Pro
  14. 40.2 GPT-5 mini
  15. 40.0 GPT-4o
  16. 39.8 Claude Sonnet 4.5
  17. 39.0 Claude 3.7 Sonnet
  18. 38.0 Gemini 2.5 Flash
  19. 38.0 Claude Opus 4
  20. 37.2 GPT-5 nano
  21. 37.0 o1
  22. 35.0 Claude Opus 4.1
  23. 34.0 GPT-4o mini
  24. 34.0 Claude Sonnet 4
  25. 33.0 Claude 3.5 Sonnet
  26. 33.0 Claude 3.5 Sonnet (October 2024)
  27. 30.0 Gemini 2.5 Flash-Lite
  1. 1 Gemini 3 Pro 91%
  2. 2 GPT-5.2 xhigh 84%
  3. 3 Gemini 3 Flash 72.6%
  4. 4 GPT-5 high 66%
  5. 5 GPT-5.1 high 58.7%
  6. 6 o4-mini medium 57.5%
  7. 7 o3 medium 52%
  8. 8 Gemini 2.5 Pro Mar 2025 48%
  9. 9 Gemini 2.5 Pro Jun 2025 46.4%
  10. 10 Gemini 2.5 Flash Sep 2025 46.2%
  11. 11 GPT-4.5 45%
  12. 12 gemini-robotics-er-1.5-preview 40.8%
  13. 13 Gemini 2.5 Pro May 2025 40.5%
  14. 14 GPT-5 mini medium 40.2%
  15. 15 GPT-4o Nov 2024 40%
  16. 16 Claude Sonnet 4.5 32k thinking 39.8%
  17. 17 Claude 3.7 Sonnet 39%
  18. 18 Gemini 2.5 Flash Apr 2025 38%
  19. 19 Claude Opus 4 16K 38%
  20. 20 GPT-5 nano high 37.2%
  21. 21 o1 medium 37%
  22. 22 Claude Opus 4.1 35%
  23. 23 GPT-4o mini 34%
  24. 24 Claude Sonnet 4 32K 34%
  25. 25 Claude 3.5 Sonnet Jun 2024 33%
  26. 26 Claude 3.5 Sonnet (October 2024) 33%
  27. 27 Gemini 2.5 Flash-Lite Sep 2025 30%

Source: VPCT leaderboard, via Epoch AI (CC BY 4.0)

30 models. Scroll sideways for the rest.

  1. 91.0 Gemini 3 Pro
  2. 84.0 GPT-5.2
  3. 76.0 Seed2.0 Pro
  4. 72.6 Gemini 3 Flash
  5. 66.0 GPT-5
  6. 61.0 Seed1.8
  7. 58.7 GPT-5.1
  8. 57.5 o4-mini
  9. 52.0 o3
  10. 48.0 Gemini 2.5 Pro
  11. 46.4 Gemini 2.5 Pro
  12. 46.2 Gemini 2.5 Flash
  13. 45.0 GPT-4.5
  14. 40.8 gemini-robotics-er-1.5-preview
  15. 40.5 Gemini 2.5 Pro
  16. 40.2 GPT-5 mini
  17. 40.0 GPT-4o
  18. 40.0 Claude Opus 4.5
  19. 39.8 Claude Sonnet 4.5
  20. 39.0 Claude 3.7 Sonnet
  21. 38.0 Gemini 2.5 Flash
  22. 38.0 Claude Opus 4
  23. 37.2 GPT-5 nano
  24. 37.0 o1
  25. 35.0 Claude Opus 4.1
  26. 34.0 GPT-4o mini
  27. 34.0 Claude Sonnet 4
  28. 33.0 Claude 3.5 Sonnet
  29. 33.0 Claude 3.5 Sonnet (October 2024)
  30. 30.0 Gemini 2.5 Flash-Lite
  1. 1 Gemini 3 Pro 91%
  2. 2 GPT-5.2 xhigh 84%
  3. 3 Seed2.0 Pro 76%
  4. 4 Gemini 3 Flash 72.6%
  5. 5 GPT-5 high 66%
  6. 6 Seed1.8 61%
  7. 7 GPT-5.1 high 58.7%
  8. 8 o4-mini medium 57.5%
  9. 9 o3 medium 52%
  10. 10 Gemini 2.5 Pro Mar 2025 48%
  11. 11 Gemini 2.5 Pro Jun 2025 46.4%
  12. 12 Gemini 2.5 Flash Sep 2025 46.2%
  13. 13 GPT-4.5 45%
  14. 14 gemini-robotics-er-1.5-preview 40.8%
  15. 15 Gemini 2.5 Pro May 2025 40.5%
  16. 16 GPT-5 mini medium 40.2%
  17. 17 GPT-4o Nov 2024 40%
  18. 18 Claude Opus 4.5 32k thinking 40%
  19. 19 Claude Sonnet 4.5 32k thinking 39.8%
  20. 20 Claude 3.7 Sonnet 39%
  21. 21 Gemini 2.5 Flash Apr 2025 38%
  22. 22 Claude Opus 4 16K 38%
  23. 23 GPT-5 nano high 37.2%
  24. 24 o1 medium 37%
  25. 25 Claude Opus 4.1 35%
  26. 26 GPT-4o mini 34%
  27. 27 Claude Sonnet 4 32K 34%
  28. 28 Claude 3.5 Sonnet Jun 2024 33%
  29. 29 Claude 3.5 Sonnet (October 2024) 33%
  30. 30 Gemini 2.5 Flash-Lite Sep 2025 30%

Sources: VPCT leaderboard, via Epoch AI (CC BY 4.0) and labs’ model cards

#ModelScoreSettingDateSource
1 Gemini 3 Pro 91% Nov 2025 VPCT leaderboard, via Epoch AI
2 GPT-5.2 84% xhigh Dec 2025 VPCT leaderboard, via Epoch AI
3 Gemini 3 Flash 72.6% Dec 2025 VPCT leaderboard, via Epoch AI
4 GPT-5 66% high Aug 2025 VPCT leaderboard, via Epoch AI
5 GPT-5.1 58.7% high Nov 2025 VPCT leaderboard, via Epoch AI
6 o4-mini 57.5% medium Apr 2025 VPCT leaderboard, via Epoch AI
7 o3 52% medium Apr 2025 VPCT leaderboard, via Epoch AI
8 Gemini 2.5 Pro Mar 2025 48% Mar 2025 VPCT leaderboard, via Epoch AI
9 Gemini 2.5 Pro Jun 2025 46.4% Jun 2025 VPCT leaderboard, via Epoch AI
10 Gemini 2.5 Flash Sep 2025 46.2% Sep 2025 VPCT leaderboard, via Epoch AI
11 GPT-4.5 45% Feb 2025 VPCT leaderboard, via Epoch AI
12 gemini-robotics-er-1.5-preview 40.8% Sep 2025 VPCT leaderboard, via Epoch AI
13 Gemini 2.5 Pro May 2025 40.5% May 2025 VPCT leaderboard, via Epoch AI
14 GPT-5 mini 40.2% medium Aug 2025 VPCT leaderboard, via Epoch AI
15 GPT-4o Nov 2024 40% Nov 2024 VPCT leaderboard, via Epoch AI
16 Claude Sonnet 4.5 39.8% 32k thinking Sep 2025 VPCT leaderboard, via Epoch AI
17 Claude 3.7 Sonnet 39% Feb 2025 VPCT leaderboard, via Epoch AI
18 Gemini 2.5 Flash Apr 2025 38% Apr 2025 VPCT leaderboard, via Epoch AI
19 Claude Opus 4 38% 16K May 2025 VPCT leaderboard, via Epoch AI
20 GPT-5 nano 37.2% high Aug 2025 VPCT leaderboard, via Epoch AI
21 o1 37% medium Dec 2024 VPCT leaderboard, via Epoch AI
22 Claude Opus 4.1 35% Aug 2025 VPCT leaderboard, via Epoch AI
23 GPT-4o mini 34% Jul 2024 VPCT leaderboard, via Epoch AI
24 Claude Sonnet 4 34% 32K May 2025 VPCT leaderboard, via Epoch AI
25 Claude 3.5 Sonnet 33% Jun 2024 Jun 2024 VPCT leaderboard, via Epoch AI
26 Claude 3.5 Sonnet (October 2024) 33% Oct 2024 VPCT leaderboard, via Epoch AI
27 Gemini 2.5 Flash-Lite Sep 2025 30% Sep 2025 VPCT leaderboard, via Epoch AI
#ModelScoreSettingDateSource
1 Gemini 3 Pro 91% Nov 2025 VPCT leaderboard, via Epoch AI
2 GPT-5.2 84% xhigh Dec 2025 VPCT leaderboard, via Epoch AI
3 Seed2.0 Pro 76% Feb 2026 Lab model card
4 Gemini 3 Flash 72.6% Dec 2025 VPCT leaderboard, via Epoch AI
5 GPT-5 66% high Aug 2025 VPCT leaderboard, via Epoch AI
6 Seed1.8 61% Dec 2025 Lab model card
7 GPT-5.1 58.7% high Nov 2025 VPCT leaderboard, via Epoch AI
8 o4-mini 57.5% medium Apr 2025 VPCT leaderboard, via Epoch AI
9 o3 52% medium Apr 2025 VPCT leaderboard, via Epoch AI
10 Gemini 2.5 Pro Mar 2025 48% Mar 2025 VPCT leaderboard, via Epoch AI
11 Gemini 2.5 Pro Jun 2025 46.4% Jun 2025 VPCT leaderboard, via Epoch AI
12 Gemini 2.5 Flash Sep 2025 46.2% Sep 2025 VPCT leaderboard, via Epoch AI
13 GPT-4.5 45% Feb 2025 VPCT leaderboard, via Epoch AI
14 gemini-robotics-er-1.5-preview 40.8% Sep 2025 VPCT leaderboard, via Epoch AI
15 Gemini 2.5 Pro May 2025 40.5% May 2025 VPCT leaderboard, via Epoch AI
16 GPT-5 mini 40.2% medium Aug 2025 VPCT leaderboard, via Epoch AI
17 GPT-4o Nov 2024 40% Nov 2024 VPCT leaderboard, via Epoch AI
18 Claude Opus 4.5 40% 32k thinking Nov 2025 Collected by Epoch AI, source not named
19 Claude Sonnet 4.5 39.8% 32k thinking Sep 2025 VPCT leaderboard, via Epoch AI
20 Claude 3.7 Sonnet 39% Feb 2025 VPCT leaderboard, via Epoch AI
21 Gemini 2.5 Flash Apr 2025 38% Apr 2025 VPCT leaderboard, via Epoch AI
22 Claude Opus 4 38% 16K May 2025 VPCT leaderboard, via Epoch AI
23 GPT-5 nano 37.2% high Aug 2025 VPCT leaderboard, via Epoch AI
24 o1 37% medium Dec 2024 VPCT leaderboard, via Epoch AI
25 Claude Opus 4.1 35% Aug 2025 VPCT leaderboard, via Epoch AI
26 GPT-4o mini 34% Jul 2024 VPCT leaderboard, via Epoch AI
27 Claude Sonnet 4 34% 32K May 2025 VPCT leaderboard, via Epoch AI
28 Claude 3.5 Sonnet 33% Jun 2024 Jun 2024 VPCT leaderboard, via Epoch AI
29 Claude 3.5 Sonnet (October 2024) 33% Oct 2024 VPCT leaderboard, via Epoch AI
30 Gemini 2.5 Flash-Lite Sep 2025 30% Sep 2025 VPCT leaderboard, via Epoch AI

Labs report scores for 2 models in their cards. Epoch AI collected claims for 1 model from labs’ reports and papers. Switch the scores above to rank them together.

6 models have held number one on VPCT leaderboard since its release in January 2025 6 models have held number one with claims included since its release in January 2025

Lab claims count from the day the card came out, independent runs from the model’s release. We draw one row each for VPCT leaderboard, which Epoch AI copied and what labs claim in their cards.

OpenAI Google DeepMind ByteDance Seed
VPCT leaderboard, via Epoch AI Passed 85% in Nov 2025, by Gemini 3 Pro
Gemini 3 Pro 91%
Lab claims Not tested since Mar 2026
Seed2.0 Pro 76%

Sources: VPCT leaderboard, via Epoch AI (CC BY 4.0) and labs’ model cards

Gemini 3 Pro on VPCT leaderboard, via Epoch AI passed 85% in November 2025

All the scores we have, one dot each, so a model tested at several settings appears more than once.

Reported by the lab in its model card On another leaderboard, collected by Epoch AI A claim Epoch AI collected from a report or paper Best so far
0% 20% 40% 60% 80% 100% 2024 2025 2026 Near the ceiling (85%) Random guessing Gemini 3 Pro GPT-5 o4-mini Gemini 2.5 Pro GPT-4o GPT-4o mini 0% 20% 40% 60% 80% 100% 2024 2025 2026 Near the ceiling (85%) Random guessing Gemini 3 Pro GPT-5

Sources: Other boards, via Epoch AI (CC BY 4.0), labs’ model cards and claims collected by Epoch AI (CC BY 4.0)

Labs headline it less often now, last in December 2025

Launch posts from every lab that has listed it, one tick per post. Tall colored ticks include it, and short gray ones leave it out.

ByteDance Seed
DateLaunch postScoreAs the post writes it
Dec 18, 2025 Seed1.8 61 VPCT

2 system cards report it, from 1 lab

Besides launch posts, we read the system cards, model cards and technical reports labs publish with a release. They report far more, including safety and frontier-risk tests.

DateDocumentScoreAs the document writes itSection
Mar 21, 2026 Seed1.8 model card 61 VPCT (Table 2) Capabilities
Feb 14, 2026 Seed2.0 Pro model card 76 VPCT (Table 8) Capabilities

More in World models

All 39 in World models
  • AgentWorldBench In use Language world models predict the next environment observation after an agent action, in 7 domains from MCP and search to Android and desktop OS. A judge model scores each prediction against the real observation on five dimensions. GPT-5.4 58.3
  • IntPhys 2 In use Models watch simulated videos and judge whether each event is physically possible, testing object permanence, immutability, continuity and solidity. Scored by accuracy. Muse Spark Open 68.3%
  • QwenWorldBench Fading Models act as world models that simulate agent environments in 7 domains, such as terminal, web, and Android. A rubric judge scores five dimensions. Qwen3.7-Plus 62.1%
  • ContPhy Fading Models infer physical properties such as mass and density of fluids and soft bodies from videos and predict their dynamics. Seed2.0 Pro 67.4%
  • RBench Only in system cards 650 image+text prompts for robot video generation (tasks and embodiments). MLLM judge scores task completion, physical plausibility, stability. Wan 2.6 60.7%
  • PAI-Bench Only in system cards Physical-AI video generation from 1,044 prompts in six domains, VLM-QA plus quality metrics. Companion control and understanding tracks. Cosmos3-Nano 83.7%