Benchmark Almanac

SnorkelSpatial

GPT-5.4 99% leads since March 2026
Not in model cards

Models track particles on a 2D grid through 10 to 200 text-described moves, then report final positions and orientations. The 330 samples test allocentric and egocentric spatial reasoning.

Text-only input with structured JSON answers checked against programmatic ground truth. The board is archived.

Made by
Snorkel AI
Released
October 2025
Kind
Language reasoning
Level
School
Scored
Checked automatically
Answers
Short answer
Tasks
Kept private
Strongest model tested
GPT-5.4 #27 on the capability index
Last model added
March 2026
Scores refreshed
November 12th, 2026
33
Entries on its own leaderboard, 27 models we track

GPT-5.4 leads with 99% among models released by March 2026

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.

The newest model with a score here came out in March 2026, and nobody has tested a newer one. Among the 20 highest-ranked models on the capability index, these have no score yet:

33 entries. Scroll sideways for the rest.

  1. 99.0 GPT-5.4
  2. 84.8 Grok 4 Fast
  3. 76.7 o3
  4. 73.9 GPT-5
  5. 52.7 gpt-oss-120b
  6. 45.5 GPT-5 mini
  7. 45.1 Claude Opus 4.1
  8. 44.2 Magistral Medium 1.2
  9. 40.3 Claude Opus 4
  10. 37.9 o3-mini
  11. 33.3 Claude Sonnet 4
  12. 26.7 GPT-5 nano
  13. 21.5 Claude 3.7 Sonnet
  14. 18.8 Gemini 2.5 Flash [snapshot unspecified]
  15. 15.4 Llama 4 Scout
  16. 15.2 Gemini 2.5 Pro [snapshot unspecified]
  17. 14.8 Mistral Large [snapshot unspecified]
  18. 14.8 o4-mini
  19. 14.8 GPT-5 Chat
  20. 14.6 Amazon Nova Micro
  21. 14.6 Llama 3.3 70B
  22. 14.6 GPT-4.1
  23. 14.6 Mistral Medium 3.1
  24. 14.2 Command R+
  25. 14.2 Amazon Nova Premier
  26. 13.9 Qwen 3 235B [snapshot unspecified]
  27. 13.6 Codestral [snapshot unspecified]
  28. 13.3 Amazon Nova Lite
  29. 12.7 Grok 3
  30. 12.4 Magistral Medium [snapshot unspecified]
  31. 12.1 Amazon Nova Pro
  32. 12.1 Llama 4 Maverick
  33. 11.8 Command R (03-2024)
  1. 1 GPT-5.4 99%
  2. 2 Grok 4 Fast reasoning 84.8%
  3. 3 o3 76.7%
  4. 4 GPT-5 73.9%
  5. 5 gpt-oss-120b 52.7%
  6. 6 GPT-5 mini 45.5%
  7. 7 Claude Opus 4.1 45.1%
  8. 8 Magistral Medium 1.2 44.2%
  9. 9 Claude Opus 4 40.3%
  10. 10 o3-mini 37.9%
  11. 11 Claude Sonnet 4 33.3%
  12. 12 GPT-5 nano 26.7%
  13. 13 Claude 3.7 Sonnet 21.5%
  14. 14 Gemini 2.5 Flash [snapshot unspecified] 18.8%
  15. 15 Llama 4 Scout 15.4%
  16. 16 Gemini 2.5 Pro [snapshot unspecified] 15.2%
  17. 17 Mistral Large [snapshot unspecified] 14.8%
  18. 18 o4-mini 14.8%
  19. 19 GPT-5 Chat 14.8%
  20. 20 Amazon Nova Micro 14.6%
  21. 21 Llama 3.3 70B 14.6%
  22. 22 GPT-4.1 14.6%
  23. 23 Mistral Medium 3.1 14.6%
  24. 24 Command R+ 14.2%
  25. 25 Amazon Nova Premier 14.2%
  26. 26 Qwen 3 235B [snapshot unspecified] 13.9%
  27. 27 Codestral [snapshot unspecified] 13.6%
  28. 28 Amazon Nova Lite 13.3%
  29. 29 Grok 3 12.7%
  30. 30 Magistral Medium [snapshot unspecified] 12.4%
  31. 31 Amazon Nova Pro 12.1%
  32. 32 Llama 4 Maverick 12.1%
  33. 33 Command R (03-2024) 11.8%

Source: Official SnorkelSpatial leaderboard

#EntryScoreSettingDateSource
1 GPT-5.4 99% Mar 2026 Official leaderboard
2 Grok 4 Fast 84.8% reasoning Sep 2025 Official leaderboard
3 o3 76.7% Apr 2025 Official leaderboard
4 GPT-5 73.9% Aug 2025 Official leaderboard
5 gpt-oss-120b 52.7% Aug 2025 Official leaderboard
6 GPT-5 mini 45.5% Aug 2025 Official leaderboard
7 Claude Opus 4.1 45.1% Aug 2025 Official leaderboard
8 Magistral Medium 1.2 44.2% Sep 2025 Official leaderboard
9 Claude Opus 4 40.3% May 2025 Official leaderboard
10 o3-mini 37.9% Jan 2025 Official leaderboard
11 Claude Sonnet 4 33.3% May 2025 Official leaderboard
12 GPT-5 nano 26.7% Aug 2025 Official leaderboard
13 Claude 3.7 Sonnet 21.5% Feb 2025 Official leaderboard
14 Gemini 2.5 Flash [snapshot unspecified] 18.8% Official leaderboard
15 Llama 4 Scout 15.4% Apr 2025 Official leaderboard
16 Gemini 2.5 Pro [snapshot unspecified] 15.2% Official leaderboard
17 Mistral Large [snapshot unspecified] 14.8% Official leaderboard
18 o4-mini 14.8% Apr 2025 Official leaderboard
19 GPT-5 Chat 14.8% Aug 2025 Official leaderboard
20 Amazon Nova Micro 14.6% Dec 2024 Official leaderboard
21 Llama 3.3 70B 14.6% Dec 2024 Official leaderboard
22 GPT-4.1 14.6% Apr 2025 Official leaderboard
23 Mistral Medium 3.1 14.6% Aug 2025 Official leaderboard
24 Command R+ 14.2% Aug 2024 Official leaderboard
25 Amazon Nova Premier 14.2% Apr 2025 Official leaderboard
26 Qwen 3 235B [snapshot unspecified] 13.9% Official leaderboard
27 Codestral [snapshot unspecified] 13.6% Official leaderboard
28 Amazon Nova Lite 13.3% Dec 2024 Official leaderboard
29 Grok 3 12.7% Apr 2025 Official leaderboard
30 Magistral Medium [snapshot unspecified] 12.4% Official leaderboard
31 Amazon Nova Pro 12.1% Dec 2024 Official leaderboard
32 Llama 4 Maverick 12.1% Apr 2025 Official leaderboard
33 Command R (03-2024) 11.8% Mar 2024 Official leaderboard

2 models have held number one on the official leaderboard since its release in October 2025

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

xAI OpenAI
Own leaderboard Passed 85% in Mar 2026, by GPT-5.4
GPT-5.4 99%

Source: Official SnorkelSpatial leaderboard

GPT-5.4 on the official leaderboard passed 85% in March 2026

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 Best so far
0% 20% 40% 60% 80% 100% 2024 2025 2026 Near the ceiling (85%) GPT-5.4 Grok 4 Fast o3 o3-mini Amazon Nova Micro Command R (03-2024) 0% 20% 40% 60% 80% 100% 2024 2025 2026 Near the ceiling (85%) GPT-5.4 Grok 4 Fast

Source: Official SnorkelSpatial 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 Knowledge

All 121 in Knowledge
  • MMLU-Pro In use Harder 10-option version of MMLU with reasoning-focused questions across 14 subjects. Intern-S2-397B 89.8%
  • SuperGPQA In use 26,529 graduate-level multiple-choice questions across 285 disciplines. Accuracy. Qwen 3.6 Max (Preview) 73.9%
  • WideSearch In use 200 bilingual broad info-seeking tasks: agent collects many atomic facts into a table. Scored by success rate and row/item F1. Tencent Hy4 preview 83.9%
  • BBEH micro average In use Models solve harder replacements for each of the 23 BIG-Bench Hard tasks, 4,520 examples in total. The score is the share of all examples answered correctly. Gemma 4 31B IT 74.4%
  • KMMLU-Pro In use Models answer multiple-choice questions taken from Korean National Professional Licensure exams. It tests professional knowledge used in Korean industry. Solar Pro 4 79.2%
  • HLE-Verified In use Verified and revised version of Humanity's Last Exam (1,811 items) with uncertain items removed. Seed2.0 Pro 73.6%