Benchmark Almanac

VoxelBench

GPT-6 Astra 2,726 leads since September 2026
Not in model cards

Language models turn text or image prompts into voxel structures. Voters compare anonymous builds from the same prompt, and Glicko-2 ratings rank the models.

Separate text-prompt and image-prompt boards. A model needs 50 votes to appear. Different from MineBench.

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 Astra #3 on the capability index
Last model added
October 2026
Scores refreshed
November 12th, 2026
57
Entries on its own leaderboard, 56 models we track

GPT-6 Astra leads with 2,726

Best score per entry from the benchmark's own leaderboard, read on November 12th, 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.

57 entries. Scroll sideways for the rest. Bars start at 1,000.

  1. 2,726 GPT-6 Astra
  2. 2,579 Claude Opus 5.5
  3. 2,532 GPT-6.1 Sol
  4. 2,368 GPT-6 Sol
  5. 2,314 GPT-6.1 Terra
  6. 2,294 Claude Opus 5
  7. 2,252 GPT-5.6 Sol
  8. 2,177 Kimi K3.1 Pro
  9. 2,144 Gemini 4 Argon
  10. 2,112 Claude Fable 5
  11. 2,077 Grok 4.6
  12. 2,037 DeepSeek V4.1 Flash 1017
  13. 1,963 GPT-5.5 Pro
  14. 1,921 GPT-5.6 Terra
  15. 1,915 GPT-5.5
  16. 1,907 Kimi K3
  17. 1,881 DeepSeek V4.1 Flash
  18. 1,876 Gemini 3.7 Flash
  19. 1,865 space-bunny-alpha
  20. 1,828 Muse Wonder
  21. 1,818 GPT-6 Luna
  22. 1,817 GPT-5.6 Luna
  23. 1,816 MiniMax-M3.1-Flash
  24. 1,810 Gemini 3.8 Flash
  25. 1,785 Gemini 3.6 Flash
  26. 1,785 GLM-5.3
  27. 1,757 Qwen 3.8 Max
  28. 1,742 GLM-5.3-Flash
  29. 1,741 Grok 4.5
  30. 1,731 Claude Haiku 5.5
  31. 1,646 Gemini 3.5 Flash
  32. 1,624 DeepSeek V4 Pro 0813
  33. 1,619 Gemini 3.1 Pro
  34. 1,609 Qwen3.7-Max
  35. 1,589 Muse Spark 1.3
  36. 1,586 Claude Opus 4.8
  37. 1,585 GLM-5.2
  38. 1,540 Claude Sonnet 5
  39. 1,534 Muse Spark 1.1
  40. 1,461 Gemini 2.5 Deep Think
  41. 1,446 Gemini 3 Deep Think
  42. 1,444 DeepSeek V4 Flash 0731
  43. 1,443 GLM-5.1
  44. 1,424 GPT-5.4
  45. 1,419 Claude Opus 4.6
  46. 1,408 MiniMax-M3
  47. 1,406 GPT-5.2
  48. 1,393 Claude Opus 4.7
  49. 1,357 Claude Opus 4.5
  50. 1,355 Gemini 3 Pro
  51. 1,313 GPT-5
  52. 1,289 DeepSeek-V4-Pro
  53. 1,286 Gemini 3 Flash
  54. 1,282 Gemini 2.5 Pro
  55. 1,263 Qwen 3.6 Plus
  56. 1,247 Claude Sonnet 4.5
  57. 1,199 GPT-5 Pro

Bars start at 1,000.

  1. 1 GPT-6 Astra max 2,726
  2. 2 Claude Opus 5.5 max 2,579
  3. 3 GPT-6.1 Sol max 2,532
  4. 4 GPT-6 Sol max 2,368
  5. 5 GPT-6.1 Terra max 2,314
  6. 6 Claude Opus 5 max 2,294
  7. 7 GPT-5.6 Sol max 2,252
  8. 8 Kimi K3.1 Pro 2,177
  9. 9 Gemini 4 Argon max 2,144
  10. 10 Claude Fable 5 max 2,112
  11. 11 Grok 4.6 xhigh 2,077
  12. 12 DeepSeek V4.1 Flash 1017 2,037
  13. 13 GPT-5.5 Pro 1,963
  14. 14 GPT-5.6 Terra 1,921
  15. 15 GPT-5.5 xhigh 1,915
  16. 16 Kimi K3 max 1,907
  17. 17 DeepSeek V4.1 Flash 1,881
  18. 18 Gemini 3.7 Flash 1,876
  19. 19 space-bunny-alpha 1,865
  20. 20 Muse Wonder max 1,828
  21. 21 GPT-6 Luna max 1,818
  22. 22 GPT-5.6 Luna 1,817
  23. 23 MiniMax-M3.1-Flash 1,816
  24. 24 Gemini 3.8 Flash 1,810
  25. 25 Gemini 3.6 Flash 1,785
  26. 26 GLM-5.3 max 1,785
  27. 27 Qwen 3.8 Max 1,757
  28. 28 GLM-5.3-Flash max 1,742
  29. 29 Grok 4.5 1,741
  30. 30 Claude Haiku 5.5 max 1,731
  31. 31 Gemini 3.5 Flash 1,646
  32. 32 DeepSeek V4 Pro 0813 1,624
  33. 33 Gemini 3.1 Pro 1,619
  34. 34 Qwen3.7-Max 1,609
  35. 35 Muse Spark 1.3 max 1,589
  36. 36 Claude Opus 4.8 max 1,586
  37. 37 GLM-5.2 max 1,585
  38. 38 Claude Sonnet 5 xhigh 1,540
  39. 39 Muse Spark 1.1 1,534
  40. 40 Gemini 2.5 Deep Think 1,461
  41. 41 Gemini 3 Deep Think 1,446
  42. 42 DeepSeek V4 Flash 0731 1,444
  43. 43 GLM-5.1 1,443
  44. 44 GPT-5.4 xhigh 1,424
  45. 45 Claude Opus 4.6 max 1,419
  46. 46 MiniMax-M3 1,408
  47. 47 GPT-5.2 xhigh 1,406
  48. 48 Claude Opus 4.7 max 1,393
  49. 49 Claude Opus 4.5 high 1,357
  50. 50 Gemini 3 Pro 1,355
  51. 51 GPT-5 high 1,313
  52. 52 DeepSeek-V4-Pro 1,289
  53. 53 Gemini 3 Flash 1,286
  54. 54 Gemini 2.5 Pro Jun 2025 1,282
  55. 55 Qwen 3.6 Plus 1,263
  56. 56 Claude Sonnet 4.5 32k thinking 1,247
  57. 57 GPT-5 Pro 1,199

Source: Official VoxelBench leaderboard

#EntryScoreSettingDateSource
1 GPT-6 Astra 2,726 max Sep 2026 Official leaderboard
2 Claude Opus 5.5 2,579 max Sep 2026 Official leaderboard
3 GPT-6.1 Sol 2,532 max Sep 2026 Official leaderboard
4 GPT-6 Sol 2,368 max Sep 2026 Official leaderboard
5 GPT-6.1 Terra 2,314 max Oct 2026 Official leaderboard
6 Claude Opus 5 2,294 max Jul 2026 Official leaderboard
7 GPT-5.6 Sol 2,252 max Jul 2026 Official leaderboard
8 Kimi K3.1 Pro 2,177 Oct 2026 Official leaderboard
9 Gemini 4 Argon 2,144 max Sep 2026 Official leaderboard
10 Claude Fable 5 2,112 max Jun 2026 Official leaderboard
11 Grok 4.6 2,077 xhigh Aug 2026 Official leaderboard
12 DeepSeek V4.1 Flash 1017 2,037 Oct 2026 Official leaderboard
13 GPT-5.5 Pro 1,963 Apr 2026 Official leaderboard
14 GPT-5.6 Terra 1,921 Jul 2026 Official leaderboard
15 GPT-5.5 1,915 xhigh Apr 2026 Official leaderboard
16 Kimi K3 1,907 max Jul 2026 Official leaderboard
17 DeepSeek V4.1 Flash 1,881 Sep 2026 Official leaderboard
18 Gemini 3.7 Flash 1,876 Aug 2026 Official leaderboard
19 space-bunny-alpha 1,865 Official leaderboard
20 Muse Wonder 1,828 max Oct 2026 Official leaderboard
21 GPT-6 Luna 1,818 max Sep 2026 Official leaderboard
22 GPT-5.6 Luna 1,817 Jul 2026 Official leaderboard
23 MiniMax-M3.1-Flash 1,816 Oct 2026 Official leaderboard
24 Gemini 3.8 Flash 1,810 Sep 2026 Official leaderboard
25 Gemini 3.6 Flash 1,785 Jul 2026 Official leaderboard
26 GLM-5.3 1,785 max Aug 2026 Official leaderboard
27 Qwen 3.8 Max 1,757 Aug 2026 Official leaderboard
28 GLM-5.3-Flash 1,742 max Aug 2026 Official leaderboard
29 Grok 4.5 1,741 Jul 2026 Official leaderboard
30 Claude Haiku 5.5 1,731 max Oct 2026 Official leaderboard
31 Gemini 3.5 Flash 1,646 May 2026 Official leaderboard
32 DeepSeek V4 Pro 0813 1,624 Aug 2026 Official leaderboard
33 Gemini 3.1 Pro 1,619 Feb 2026 Official leaderboard
34 Qwen3.7-Max 1,609 May 2026 Official leaderboard
35 Muse Spark 1.3 1,589 max Sep 2026 Official leaderboard
36 Claude Opus 4.8 1,586 max May 2026 Official leaderboard
37 GLM-5.2 1,585 max Jun 2026 Official leaderboard
38 Claude Sonnet 5 1,540 xhigh Jun 2026 Official leaderboard
39 Muse Spark 1.1 1,534 Jul 2026 Official leaderboard
40 Gemini 2.5 Deep Think 1,461 Aug 2025 Official leaderboard
41 Gemini 3 Deep Think 1,446 Feb 2026 Official leaderboard
42 DeepSeek V4 Flash 0731 1,444 Jul 2026 Official leaderboard
43 GLM-5.1 1,443 Apr 2026 Official leaderboard
44 GPT-5.4 1,424 xhigh Mar 2026 Official leaderboard
45 Claude Opus 4.6 1,419 max Feb 2026 Official leaderboard
46 MiniMax-M3 1,408 Jun 2026 Official leaderboard
47 GPT-5.2 1,406 xhigh Dec 2025 Official leaderboard
48 Claude Opus 4.7 1,393 max Apr 2026 Official leaderboard
49 Claude Opus 4.5 1,357 high Nov 2025 Official leaderboard
50 Gemini 3 Pro 1,355 Nov 2025 Official leaderboard
51 GPT-5 1,313 high Aug 2025 Official leaderboard
52 DeepSeek-V4-Pro 1,289 Apr 2026 Official leaderboard
53 Gemini 3 Flash 1,286 Dec 2025 Official leaderboard
54 Gemini 2.5 Pro Jun 2025 1,282 Jun 2025 Official leaderboard
55 Qwen 3.6 Plus 1,263 Mar 2026 Official leaderboard
56 Claude Sonnet 4.5 1,247 32k thinking Sep 2025 Official leaderboard
57 GPT-5 Pro 1,199 Oct 2025 Official leaderboard

7 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 OpenAI Anthropic
Own leaderboard 7 leaders
GPT-6 Astra 2,726

Source: Official VoxelBench 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 3,000 2025 2026 2027 1,000 1,500 2,000 2,500 3,000 2025 2026 2027

Source: Official VoxelBench 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%