Benchmark Almanac

VoxelBench Image

GPT-6.1 Sol 2,264 leads since September 2026
Not in model cards

Models turn image prompts into voxel structures. Human comparisons determine Glicko-2 ratings in a separate image-prompt vote pool.

The site requires at least 50 votes. Text-prompt ratings belong to the separate VoxelBench Text board.

Made by
VoxelBench
Released
September 2025
Kind
Spatial reasoning and counting
Level
Everyday
Scored
Ranked by votes
Answers
Open-ended
Tasks
All public
Strongest model tested
GPT-6.1 Sol #5 on the capability index
Last model added
September 2026
Scores refreshed
November 12th, 2026
13
Models scored on its own leaderboard

GPT-6.1 Sol leads with 2,264

Best score per model from the benchmark's own leaderboard, read on November 12th, 2026. Where a model ran with several settings or harnesses, we keep its best.

13 models. Scroll sideways for the rest. Bars start at 500.

  1. 2,264 GPT-6.1 Sol
  2. 1,535 Gemini 3.1 Pro
  3. 1,527 GPT-5.5
  4. 1,467 Claude Opus 4.6
  5. 1,421 GPT-5.4
  6. 1,411 Gemini 3 Pro
  7. 1,360 GPT-5.2
  8. 1,320 Claude Opus 4.5
  9. 1,266 Claude Opus 4.7
  10. 1,199 Gemini 2.5 Pro
  11. 1,168 GPT-5
  12. 1,000 Gemini 2.5 Deep Think
  13. 1,000 Gemini 3.5 Flash

Bars start at 500.

  1. 1 GPT-6.1 Sol max 2,264
  2. 2 Gemini 3.1 Pro 1,535
  3. 3 GPT-5.5 xhigh 1,527
  4. 4 Claude Opus 4.6 max 1,467
  5. 5 GPT-5.4 xhigh 1,421
  6. 6 Gemini 3 Pro 1,411
  7. 7 GPT-5.2 xhigh 1,360
  8. 8 Claude Opus 4.5 high 1,320
  9. 9 Claude Opus 4.7 max 1,266
  10. 10 Gemini 2.5 Pro Jun 2025 1,199
  11. 11 GPT-5 high 1,168
  12. 12 Gemini 2.5 Deep Think 1,000
  13. 13 Gemini 3.5 Flash 1,000

Source: Official VoxelBench Image leaderboard

#ModelScoreSettingDateSource
1 GPT-6.1 Sol 2,264 max Sep 2026 Official leaderboard
2 Gemini 3.1 Pro 1,535 Feb 2026 Official leaderboard
3 GPT-5.5 1,527 xhigh Apr 2026 Official leaderboard
4 Claude Opus 4.6 1,467 max Feb 2026 Official leaderboard
5 GPT-5.4 1,421 xhigh Mar 2026 Official leaderboard
6 Gemini 3 Pro 1,411 Nov 2025 Official leaderboard
7 GPT-5.2 1,360 xhigh Dec 2025 Official leaderboard
8 Claude Opus 4.5 1,320 high Nov 2025 Official leaderboard
9 Claude Opus 4.7 1,266 max Apr 2026 Official leaderboard
10 Gemini 2.5 Pro Jun 2025 1,199 Jun 2025 Official leaderboard
11 GPT-5 1,168 high Aug 2025 Official leaderboard
12 Gemini 2.5 Deep Think 1,000 Aug 2025 Official leaderboard
13 Gemini 3.5 Flash 1,000 May 2026 Official leaderboard

5 models have held number one on the official leaderboard since its release in September 2025

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

Google DeepMind Anthropic OpenAI
Own leaderboard 5 leaders
GPT-6.1 Sol 2,264

Source: Official VoxelBench Image leaderboard

Scores over time

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

On the benchmark’s own leaderboard
1,000 1,500 2,000 2,500 2025 2026 1,000 1,500 2,000 2,500 2025 2026

Source: Official VoxelBench Image leaderboard

No lab model card reports it yet

We read the headline results of 522 launch posts from 52 labs, and none of them list this benchmark.

More in Vision

All 281 in Vision
  • MMMU In use 11.5K college-level multimodal questions across 30 subjects. Multiple-choice and open accuracy. GPT-5.1 85.4%
  • 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%
  • 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%
  • ChartQA In use Models answer human-written and generated questions about chart images that need visual and arithmetic reasoning. Scored by relaxed accuracy with a 5 percent numeric tolerance. Command A Vision 90.9%