Benchmark Almanac

WTP-Bench

GPT-4.1 79% leads since February 2026
Not in model cards

Models identify Pokémon from official artwork and silhouettes across 1,160 entries. The scorer checks exact forms, species, and evolutionary families.

The maker publishes four accuracy measures for 27 vision models. The silhouette column measures exact form matches.

Released
February 2026
Kind
General visual questions
Level
School
Scored
Checked automatically
Answers
Short answer
Tasks
All public
Measures
Capability
Language
English
In and out
Images, Text → Text
Interaction
Single turn
Items from
Real work and use
Refresh
One fixed set
Run by
Anyone
In views
Vision
Strongest model tested
Claude Sonnet 4.5 #81 on the capability index
Last model added
October 2025
Scores refreshed
November 20th, 2026
27
Entries on its own leaderboard, 24 models we track

GPT-4.1 leads with 79% among models released by October 2025 GPT-4.1 leads with 79% among models released by October 2025

Best score per entry from the benchmark's own leaderboard, read on November 20th, 2026. Where a model ran with several settings or harnesses, we keep its best. Names in gray aren't in our model list: fine-tunes, agent systems, products and older models. They don't count toward leaders. This view shows only models we track, each at its best.

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

  1. 1 GPT-4.1 79%
  2. 2 Gemini 2.5 Pro Jun 2025 78.2%
  3. 3 Gemini 2.5 Flash Sep 2025 77.2%
  4. 4 Claude Sonnet 4.5 65.1%
  5. 5 GPT-4o mini 49.3%
  6. 6 Grok 4 Fast 46.9%
  7. 7 Claude Haiku 4.5 27.6%
  8. 8 Qwen3-VL-4B-Instruct 13.4%
  9. 9 Qwen3-VL 8B 12.8%
  10. 10 Pixtral 12B 7.16%
  11. 11 Qwen2-VL-7B-Instruct 6.47%
  12. 12 Ovis2.5-9B 5.95%
  13. 13 InternVL2.5-26B 4.14%
  14. 14 LLaVA-OneVision 7B 3.53%
  15. 15 Qwen2.5-VL-7B 3.53%
  16. 16 Florence-VL 8B 2.93%
  17. 17 Qwen2.5-VL-3B 2.84%
  18. 18 Ovis2.5-2B 2.76%
  19. 19 Qwen2-VL-2B-Instruct 2.24%
  20. 20 SmolVLM-Instruct 2.24%
  21. 21 PaliGemma-3B-mix-448 1.98%
  22. 22 PaliGemma 3B 1.81%
  23. 23 LLaVA-v1.5-13B 1.81%
  24. 24 llava-v1.5-7b 1.03%
  25. 25 InternVL2-8B 0.86%
  26. 26 Phi-3-vision-128k-instruct 0.17%
  27. 27 InternVL2-1B 0.09%

Source: Official WTP-Bench leaderboard

#EntryScoreSettingDateSource
1 GPT-4.1 79% Apr 2025 Official leaderboard
2 Gemini 2.5 Pro Jun 2025 78.2% Jun 2025 Official leaderboard
3 Gemini 2.5 Flash Sep 2025 77.2% Sep 2025 Official leaderboard
4 Claude Sonnet 4.5 65.1% Sep 2025 Official leaderboard
5 GPT-4o mini 49.3% Jul 2024 Official leaderboard
6 Grok 4 Fast 46.9% Sep 2025 Official leaderboard
7 Claude Haiku 4.5 27.6% Oct 2025 Official leaderboard
8 Qwen3-VL-4B-Instruct 13.4% Oct 2025 Official leaderboard
9 Qwen3-VL 8B open weights 12.8% Official leaderboard
10 Pixtral 12B 7.16% Sep 2024 Official leaderboard
11 Qwen2-VL-7B-Instruct 6.47% Aug 2024 Official leaderboard
12 Ovis2.5-9B 5.95% Aug 2025 Official leaderboard
13 InternVL2.5-26B 4.14% Dec 2024 Official leaderboard
14 LLaVA-OneVision 7B 3.53% Aug 2024 Official leaderboard
15 Qwen2.5-VL-7B 3.53% Jan 2025 Official leaderboard
16 Florence-VL 8B open weights 2.93% Official leaderboard
17 Qwen2.5-VL-3B 2.84% Jan 2025 Official leaderboard
18 Ovis2.5-2B 2.76% Aug 2025 Official leaderboard
19 Qwen2-VL-2B-Instruct 2.24% Aug 2024 Official leaderboard
20 SmolVLM-Instruct 2.24% Nov 2024 Official leaderboard
21 PaliGemma-3B-mix-448 1.98% May 2024 Official leaderboard
22 PaliGemma 3B open weights 1.81% Official leaderboard
23 LLaVA-v1.5-13B 1.81% Oct 2023 Official leaderboard
24 llava-v1.5-7b 1.03% Oct 2023 Official leaderboard
25 InternVL2-8B 0.86% Jul 2024 Official leaderboard
26 Phi-3-vision-128k-instruct 0.17% May 2024 Official leaderboard
27 InternVL2-1B 0.09% Jul 2024 Official leaderboard
#ModelScoreSettingDateSource
1 GPT-4.1 79% Apr 2025 Official leaderboard
2 Gemini 2.5 Pro Jun 2025 78.2% Jun 2025 Official leaderboard
3 Gemini 2.5 Flash Sep 2025 77.2% Sep 2025 Official leaderboard
4 Claude Sonnet 4.5 65.1% Sep 2025 Official leaderboard
5 GPT-4o mini 49.3% Jul 2024 Official leaderboard
6 Grok 4 Fast 46.9% Sep 2025 Official leaderboard
7 Claude Haiku 4.5 27.6% Oct 2025 Official leaderboard
8 Qwen3-VL-4B-Instruct 13.4% Oct 2025 Official leaderboard
9 Pixtral 12B 7.16% Sep 2024 Official leaderboard
10 Qwen2-VL-7B-Instruct 6.47% Aug 2024 Official leaderboard
11 Ovis2.5-9B 5.95% Aug 2025 Official leaderboard
12 InternVL2.5-26B 4.14% Dec 2024 Official leaderboard
13 LLaVA-OneVision 7B 3.53% Aug 2024 Official leaderboard
14 Qwen2.5-VL-7B 3.53% Jan 2025 Official leaderboard
15 Qwen2.5-VL-3B 2.84% Jan 2025 Official leaderboard
16 Ovis2.5-2B 2.76% Aug 2025 Official leaderboard
17 Qwen2-VL-2B-Instruct 2.24% Aug 2024 Official leaderboard
18 SmolVLM-Instruct 2.24% Nov 2024 Official leaderboard
19 PaliGemma-3B-mix-448 1.98% May 2024 Official leaderboard
20 LLaVA-v1.5-13B 1.81% Oct 2023 Official leaderboard
21 llava-v1.5-7b 1.03% Oct 2023 Official leaderboard
22 InternVL2-8B 0.86% Jul 2024 Official leaderboard
23 Phi-3-vision-128k-instruct 0.17% May 2024 Official leaderboard
24 InternVL2-1B 0.09% Jul 2024 Official leaderboard

Rankings by part

We read 3 parts from the benchmark’s own leaderboard: Ultra Ball, Great Ball and Shadow Ball. Each ranks models on one part of the test. Gemini 2.5 Pro leads 2 of them, and Claude Sonnet 4.5 leads the other one. Arrows show how far a model moved from its overall rank among the models we track.

Ultra Ball

24 models we track have a score here.

  1. 1 Gemini 2.5 Pro Jun 2025 ↑1 88.1%
  2. 2 GPT-4.1 ↓1 84.2%
  3. 3 Gemini 2.5 Flash Sep 2025 83%
  4. 4 Claude Sonnet 4.5 79.2%
  5. 5 GPT-4o mini 62%
  6. 6 Grok 4 Fast 59%
  7. 7 Claude Haiku 4.5 35.8%
  8. 8 Qwen3-VL-4B-Instruct 18.4%
  9. 9 Pixtral 12B 9.74%
  10. 10 Ovis2.5-9B ↑1 9.31%

Number one over time

Lab claims count from the day the card came out, independent runs from the model’s release.

OpenAI
Own leaderboard Not tested since Feb 2026
GPT-4.1 79%

Source: Official WTP-Bench leaderboard

Scores over time

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

Plot with other benchmarks
On the benchmark’s own leaderboard Best so far
0% 20% 40% 60% 80% 100% 2024 2025 Near the ceiling (85%) GPT-4.1 GPT-4o mini PaliGemma-3B-mix-448 LLaVA-v1.5-13B 0% 20% 40% 60% 80% 100% 2024 2025 Near the ceiling (85%) GPT-4.1 GPT-4o mini

Source: Official WTP-Bench leaderboard

No lab model card reports it yet

We read the headline results of 540 launch posts from 57 labs, and none of them list this benchmark.

More in Vision

All 636 in Vision
  • MMMU In use 11.5K college-level multimodal questions across 30 subjects. Multiple-choice and open accuracy. Qwen 3.6 Plus 86%
  • MMMU-Pro In use Harder MMMU: college-level multimodal questions with more options and vision-only inputs to remove text shortcuts. GPT-5.6 Sol 84.6%
  • CharXiv In use Chart understanding on real arXiv figures. Reasoning questions require synthesizing information from complex charts. Claude Fable 5 93.5%
  • MathVista In use Math reasoning in visual contexts (charts, geometry, puzzles) from 31 datasets. Answer accuracy. Seed2.1 Pro 90.7%
  • MATH-Vision (MathVision) In use 3,040 competition math problems with visual context. Answer accuracy. Qwen 3.8 Max 97.7%
  • MMStar In use Models answer 1,500 human-curated image questions that cannot be solved without the image, across 6 core capabilities. Scored by accuracy. Qwen 3.8 Max 85.9%

Data audit

9 of 10 checks pass, 1 warning. Look first at the Newest score check. The newest came 13 months ago, in October 2025, for Claude Haiku 4.5. Every benchmark

Catalog record 2 pass
  • Pass Task access All tasks are public.
  • Pass Confidence Checked against a primary source.
Scores 1 warning 3 pass 1 note
  • Pass Independent runs Scores from the official leaderboard (27 entries).
  • Pass Models scored 24 models we track, plus 3 other entries on its leaderboard.
  • Warning Newest score The newest came 13 months ago, in October 2025, for Claude Haiku 4.5.
  • Note Frontier models None of the top 10 has a score. The strongest tested is Claude Sonnet 4.5 (#81).
  • Pass Headroom The best is 79%, under 85%, so it still separates models.
Leaderboard import 3 pass
  • Pass Importer We read the board on sochastic.me: 24 rows for models we track and 3 other entries at the last read.
  • Pass Last read November 20th, 2026.
  • Pass Read worked The last read worked.
Consistency 1 pass 1 note
  • Note Lab claims No model has both a lab number and an independent run.
  • Pass Known issues No published audit found problems.

How we check