SkateBench v2
Gemini 3.1 Pro 96.9% leads since Mar 2026 Not in model cards Models name skateboard tricks from technical trick definitions. The score is the share named correctly.
390 trick definitions, private after v1 saturated. The board lists 28 models but stopped near Gemini 3.1 Pro (97%), so current models are missing. Release month comes from secondary reports of the maker's post.
Made by Theo Browne (t3.gg) Released Mar 2026 Trivia and open questions Tasks kept private
Last model added Mar 2026
Scores refreshed 2 Nov 2026
- Level
- Everyday Most adults, without training
- Scored
- Checked automatically A program checks the result against an answer key, unit tests or the expected final state.
- Answers
- Short answer A number, word or short string matched to a key.
21
Models scored on its own leaderboard
Gemini 3.1 Pro leads with 96.9% among models released by Mar 2026
Best score per model from the benchmark's own leaderboard, read on 2 Nov 2026. Where a model ran with several settings or harnesses, we keep its best. Every run is in the next chart.
The newest model with a score here came out in Mar 2026, and nobody has tested a newer one. Among the 20 highest-ranked models on the capability index, these have no score yet:
- GPT-6 Astra
- Claude Opus 5.5
- GPT-6.1 Terra
- GPT-6.1 Sol
- and 15 more
- 96.9 Gemini 3.1 Pro
- 81.5 GPT-5.4
- 78.5 GPT-5.4 Pro
- 77.4 GPT-5
- 76.2 Gemini 3 Pro
- 76.2 Gemini 3 Flash
- 71.8 GPT-5.1
- 71.8 GPT-5.2 Pro
- 69.0 GPT-5.2
- 66.7 GLM-5
- 64.4 Claude Opus 4.6
- 61.3 Grok 4
- 57.2 Kimi K2.5
- 50.3 Claude Opus 4.5
- 46.7
DeepSeek-V3.2 - 43.9 Kimi K2 Thinking
- 38.0 GPT-5 mini
- 35.4 Grok 4.1 Fast
- 16.4 gpt-oss-120b
- 15.1 Claude Sonnet 4.6
- 13.8
MiniMax-M2.5
- 1 Gemini 3.1 Pro 96.9%
- 2 GPT-5.4 high 81.5%
- 3 GPT-5.4 Pro thinking 78.5%
- 4 GPT-5 high 77.4%
- 5 Gemini 3 Pro 76.2%
- 6 Gemini 3 Flash low 76.2%
- 7 GPT-5.1 71.8%
- 8 GPT-5.2 Pro 71.8%
- 9 GPT-5.2 xhigh 69%
- 10 GLM-5 66.7%
- 11 Claude Opus 4.6 high 64.4%
- 12 Grok 4 61.3%
- 13 Kimi K2.5 57.2%
- 14 Claude Opus 4.5 high 50.3%
- 15 DeepSeek-V3.2 high 46.7%
- 16 Kimi K2 Thinking 43.9%
- 17 GPT-5 mini 38%
- 18 Grok 4.1 Fast 35.4%
- 19 gpt-oss-120b high 16.4%
- 20 Claude Sonnet 4.6 15.1%
- 21 MiniMax-M2.5 13.8%
| # | Model | Score | Setting | Date | Source |
|---|---|---|---|---|---|
| 1 | Gemini 3.1 Pro | 96.9% | Feb 2026 | Official leaderboard | |
| 2 | GPT-5.4 | 81.5% | high | Mar 2026 | Official leaderboard |
| 3 | GPT-5.4 Pro | 78.5% | thinking | Mar 2026 | Official leaderboard |
| 4 | GPT-5 | 77.4% | high | Aug 2025 | Official leaderboard |
| 5 | Gemini 3 Pro | 76.2% | Nov 2025 | Official leaderboard | |
| 6 | Gemini 3 Flash | 76.2% | low | Dec 2025 | Official leaderboard |
| 7 | GPT-5.1 | 71.8% | Nov 2025 | Official leaderboard | |
| 8 | GPT-5.2 Pro | 71.8% | Dec 2025 | Official leaderboard | |
| 9 | GPT-5.2 | 69% | xhigh | Dec 2025 | Official leaderboard |
| 10 | GLM-5 | 66.7% | Feb 2026 | Official leaderboard | |
| 11 | Claude Opus 4.6 | 64.4% | high | Feb 2026 | Official leaderboard |
| 12 | Grok 4 | 61.3% | Jul 2025 | Official leaderboard | |
| 13 | Kimi K2.5 | 57.2% | Jan 2026 | Official leaderboard | |
| 14 | Claude Opus 4.5 | 50.3% | high | Nov 2025 | Official leaderboard |
| 15 | DeepSeek-V3.2 | 46.7% | high | Dec 2025 | Official leaderboard |
| 16 | Kimi K2 Thinking | 43.9% | Nov 2025 | Official leaderboard | |
| 17 | GPT-5 mini | 38% | Aug 2025 | Official leaderboard | |
| 18 | Grok 4.1 Fast | 35.4% | Nov 2025 | Official leaderboard | |
| 19 | gpt-oss-120b | 16.4% | high | Aug 2025 | Official leaderboard |
| 20 | Claude Sonnet 4.6 | 15.1% | Feb 2026 | Official leaderboard | |
| 21 | MiniMax-M2.5 | 13.8% | Feb 2026 | Official leaderboard |
The best score, 96.9%, costs $0.02 a task
Runs on the cost frontier score higher than any run that's cheaper. The costs are the leaderboard's own estimates, so compare them only within this benchmark.
OpenAI DeepSeek Google DeepMind
Cost frontier
One model at each effort level
| Model | Effort | Harness | Score | Cost per task | Output tokens |
|---|---|---|---|---|---|
| Gemini 3.1 Pro | 96.9% | $0.02 | 1,718 | ||
| GPT-5.4 | high | 81.5% | $0.02 | 1,180 | |
| GPT-5.4 | xhigh | 81.3% | $0.08 | 5,336 | |
| GPT-5.4 | medium | 78.5% | $0.009 | 574 | |
| GPT-5.4 Pro | thinking | 78.5% | $0.08 | 5,134 | |
| GPT-5 | high | 77.4% | $0.02 | 2,439 | |
| Gemini 3 Pro | 76.2% | $0.02 | 1,504 | ||
| Gemini 3 Flash | low | 76.2% | $0.001 | 437 | |
| GPT-5 | 75.4% | $0.01 | 1,159 | ||
| Gemini 3 Flash | high | 75.1% | $0.005 | 1,603 | |
| GPT-5.1 | 71.8% | $0.02 | 1,854 | ||
| GPT-5.2 Pro | 71.8% | $0.07 | 386 | ||
| GPT-5.2 | xhigh | 69% | $0.03 | 2,113 | |
| GLM-5 | 66.7% | $0.006 | 2,277 | ||
| Claude Opus 4.6 | high | 64.4% | $0.02 | 593 | |
| Grok 4 | 61.3% | $0.03 | 2,260 | ||
| Kimi K2.5 | 57.2% | $0.006 | 2,064 | ||
| Claude Opus 4.5 | high | 50.3% | $0.03 | 1,219 | |
| DeepSeek-V3.2 | high | 46.7% | $0.0008 | 1,398 | |
| GPT-5.2 | high | 44.6% | $0.005 | 338 | |
| Kimi K2 Thinking | 43.9% | $0.003 | 1,260 | ||
| GPT-5.2 | 41.5% | $0.003 | 216 | ||
| GPT-5 mini | 38% | $0.002 | 784 | ||
| Grok 4.1 Fast | 35.4% | $0.001 | 1,939 | ||
| GPT-5 | minimal | 16.7% | $0.0004 | 23 | |
| gpt-oss-120b | high | 16.4% | $0.001 | 2,175 | |
| Claude Sonnet 4.6 | 15.1% | $0.01 | 655 | ||
| MiniMax-M2.5 | 13.8% | $0.0005 | 356 |
Number one over time
Lab claims count from the day the card came out, independent runs from the model’s release.
Google DeepMind
The best score passed 85% in Feb 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
No lab model card reports it yet
We read the headline results of 479 launch posts from 51 labs, and none of them list this benchmark.
More 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% - 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% - BIG-Bench Extra Hard In use Models solve harder replacements for each of the 23 BIG-Bench Hard tasks, 4,520 examples in total. The score is a harmonic mean of task accuracies. Gemma 4 31B IT 74.4%
- 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%