Visual Strategic Bench
Models identify objects, predict another player’s next action, and choose their own moves in ten games. The authors score each ability separately.
The authors publish results for 13 distinct models. Their 15 configurations include two repeat runs with reasoning disabled.
- Made by
- Zelai Xu and colleagues
- Released
- June 2025
- Kind
- Multi-agent games
- Level
- Everyday
- Scored
- Checked automatically
- Answers
- Agent task
- Tasks
- All public
- Measures
- Capability
- Language
- English
- Interaction
- Multi-agent
- Environment
- A simulator
- Items from
- Generated, Live play or votes
- Refresh
- Live
- Run by
- Anyone
- In views
- Agents, Vision, Robotics, Kept fresh
- Strongest model tested
- o3 #84 on the capability index
- Last model added
- April 2025
- Scores refreshed
- November 20th, 2026
o3 leads with 84.9% among models released by April 2025 o3 leads with 84.9% among models released by April 2025
Best score per entry from the benchmark's own leaderboard, read on November 20th, 2026. The headline is overall perception accuracy, the first of the three abilities tested. 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 April 2025, and nobody has tested a newer one. Among the 20 highest-ranked models on the capability index, these have no score yet:
- GPT-6.2 Astra
- Claude Fable 5.2
-
Gemini 4 Carbon
- GPT-6 Astra
- and 16 more
- 1 o3 84.9%
- 2 gemini-2.5-pro (snapshot unspecified) Thinking off 84.5%
- 3
UI-TARS-1.5 81.5%
- 4 Claude 3.7 Sonnet 81.1%
- 5
Qwen2.5-VL-72B 80.3%
- 6 GPT-4.1 80.3%
- 7 Qwen-VL-Max 80.2%
- 8 doubao-1-5-vision-pro 77.6%
- 9 doubao-1-5-thinking-pro 74.7%
- 10 qvq-max 74.5%
- 11
InternVL3-78B 74.1%
- 12 Grok 2 Vision (1212) 70.2%
- 13
Llama 3.2 90B 67.8%
Source: Official Visual Strategic Bench leaderboard
- 1 o3 84.9%
- 2
UI-TARS-1.5 81.5%
- 3 Claude 3.7 Sonnet 81.1%
- 4
Qwen2.5-VL-72B 80.3%
- 5 GPT-4.1 80.3%
- 6 Qwen-VL-Max 80.2%
- 7
InternVL3-78B 74.1%
- 8 Grok 2 Vision (1212) 70.2%
- 9
Llama 3.2 90B 67.8%
Source: Official Visual Strategic Bench leaderboard
| # | Entry | Score | Setting | Date | Source |
|---|---|---|---|---|---|
| 1 | o3 | 84.9% | Apr 2025 | Official leaderboard | |
| 2 | gemini-2.5-pro (snapshot unspecified) | 84.5% | Thinking off | Official leaderboard | |
| 3 | | 81.5% | Apr 2025 | Official leaderboard | |
| 4 | Claude 3.7 Sonnet | 81.1% | Feb 2025 | Official leaderboard | |
| 5 | | 80.3% | Jan 2025 | Official leaderboard | |
| 6 | GPT-4.1 | 80.3% | Apr 2025 | Official leaderboard | |
| 7 | Qwen-VL-Max | 80.2% | Jan 2024 | Official leaderboard | |
| 8 | doubao-1-5-vision-pro | 77.6% | Official leaderboard | ||
| 9 | doubao-1-5-thinking-pro | 74.7% | Official leaderboard | ||
| 10 | qvq-max | 74.5% | Official leaderboard | ||
| 11 | | 74.1% | Apr 2025 | Official leaderboard | |
| 12 | Grok 2 Vision (1212) | 70.2% | Dec 2024 | Official leaderboard | |
| 13 | | 67.8% | Sep 2024 | Official leaderboard |
| # | Model | Score | Setting | Date | Source |
|---|---|---|---|---|---|
| 1 | o3 | 84.9% | Apr 2025 | Official leaderboard | |
| 2 | | 81.5% | Apr 2025 | Official leaderboard | |
| 3 | Claude 3.7 Sonnet | 81.1% | Feb 2025 | Official leaderboard | |
| 4 | | 80.3% | Jan 2025 | Official leaderboard | |
| 5 | GPT-4.1 | 80.3% | Apr 2025 | Official leaderboard | |
| 6 | Qwen-VL-Max | 80.2% | Jan 2024 | Official leaderboard | |
| 7 | | 74.1% | Apr 2025 | Official leaderboard | |
| 8 | Grok 2 Vision (1212) | 70.2% | Dec 2024 | Official leaderboard | |
| 9 | | 67.8% | Sep 2024 | Official leaderboard |
Rankings by part
We read 33 parts from the benchmark’s own leaderboard, each ranking models on one part of the test, such as Perception, Perception: Hanabi and Perception: Overcooked. o3 leads 17 of them, and 5 other models lead the rest. Arrows show how far a model moved from its overall rank among the models we track.
Perception
9 models we track have a score here.
- 1 o3 84.9%
- 2
UI-TARS-1.5 81.5%
- 3 Claude 3.7 Sonnet 81.1%
- 4
Qwen2.5-VL-72B 80.3%
- 5 GPT-4.1 80.3%
- 6 Qwen-VL-Max 80.2%
- 7
InternVL3-78B 74.1%
- 8 Grok 2 Vision (1212) 70.2%
- 9
Llama 3.2 90B 67.8%
Perception: Hanabi
9 models we track have a score here.
- 1 o3 79.7%
- 2 Qwen-VL-Max ↑4 76.1%
- 3
Qwen2.5-VL-72B ↑1 76%
- 4 Claude 3.7 Sonnet ↓1 75.9%
- 5
UI-TARS-1.5 ↓3 75.4%
- 6 Grok 2 Vision (1212) ↑2 75.2%
- 7
InternVL3-78B 74.6%
- 8 GPT-4.1 ↓3 72.1%
- 9
Llama 3.2 90B 30.7%
Perception: Overcooked
9 models we track have a score here.
- 1
Qwen2.5-VL-72B ↑3 72.9%
- 2 o3 ↓1 69.8%
- 3 Qwen-VL-Max ↑3 68.2%
- 4 Claude 3.7 Sonnet ↓1 62.8%
- 5 GPT-4.1 62%
- 6
UI-TARS-1.5 ↓4 59.8%
- 7
Llama 3.2 90B ↑2 58.6%
- 8 Grok 2 Vision (1212) 46.8%
- 9
InternVL3-78B ↓2 43.6%
Perception: KAZ
9 models we track have a score here.
- 1 Claude 3.7 Sonnet ↑2 92.9%
- 2 GPT-4.1 ↑3 92.6%
- 3 o3 ↓2 90.1%
- 4
Llama 3.2 90B ↑5 87.1%
- 5
Qwen2.5-VL-72B ↓1 85.5%
- 6 Qwen-VL-Max 85.2%
- 7
UI-TARS-1.5 ↓5 83%
- 8
InternVL3-78B ↓1 63.8%
- 9 Grok 2 Vision (1212) ↓1 59.8%
Perception: Board
9 models we track have a score here.
- 1
UI-TARS-1.5 ↑1 98%
- 2 o3 ↓1 97.2%
- 3 Qwen-VL-Max ↑3 81.2%
- 4 Grok 2 Vision (1212) ↑4 80.3%
- 5 Claude 3.7 Sonnet ↓2 79%
- 6
Qwen2.5-VL-72B ↓2 75.1%
- 7 GPT-4.1 ↓2 67%
- 8
InternVL3-78B ↓1 64.3%
- 9
Llama 3.2 90B 59.7%
Perception: Poker
9 models we track have a score here.
- 1
Qwen2.5-VL-72B ↑3 100%
- 2 GPT-4.1 ↑3 100%
- 3
UI-TARS-1.5 ↓1 100%
- 4 Claude 3.7 Sonnet ↓1 99.6%
- 5 Qwen-VL-Max ↑1 99.2%
- 6
InternVL3-78B ↑1 99.2%
- 7 o3 ↓6 99.1%
- 8
Llama 3.2 90B ↑1 81.6%
- 9 Grok 2 Vision (1212) ↓1 59.5%
Perception: Pong
9 models we track have a score here.
- 1 GPT-4.1 ↑4 75.4%
- 2
UI-TARS-1.5 74.4%
- 3 Grok 2 Vision (1212) ↑5 71%
- 4 Claude 3.7 Sonnet ↓1 70.1%
- 5 Qwen-VL-Max ↑1 69%
- 6
InternVL3-78B ↑1 66.7%
- 7
Qwen2.5-VL-72B ↓3 65.2%
- 8 o3 ↓7 64.6%
- 9
Llama 3.2 90B 59.5%
Perception: MPE
9 models we track have a score here.
- 1 GPT-4.1 ↑4 98.9%
- 2 o3 ↓1 97.5%
- 3
Llama 3.2 90B ↑6 94.3%
- 4
UI-TARS-1.5 ↓2 92.2%
- 5 Claude 3.7 Sonnet ↓2 90.6%
- 6
InternVL3-78B ↑1 89.2%
- 7 Qwen-VL-Max ↓1 88.2%
- 8
Qwen2.5-VL-72B ↓4 87.7%
- 9 Grok 2 Vision (1212) ↓1 78.7%
Perception: Dilemma
9 models we track have a score here.
- 1 o3 85.5%
- 2 Claude 3.7 Sonnet ↑1 82.6%
- 3
InternVL3-78B ↑4 81.1%
- 4
Qwen2.5-VL-72B 79.8%
- 5 Qwen-VL-Max ↑1 78.4%
- 6 GPT-4.1 ↓1 76.7%
- 7 Grok 2 Vision (1212) ↑1 76.4%
- 8
UI-TARS-1.5 ↓6 76%
- 9
Llama 3.2 90B 68.5%
Perception: Hunt
9 models we track have a score here.
- 1 Claude 3.7 Sonnet ↑2 80.4%
- 2 o3 ↓1 80.2%
- 3
Qwen2.5-VL-72B ↑1 79%
- 4 GPT-4.1 ↑1 76.8%
- 5
InternVL3-78B ↑2 76.7%
- 6 Qwen-VL-Max 76.4%
- 7
UI-TARS-1.5 ↓5 74.9%
- 8 Grok 2 Vision (1212) 73.3%
- 9
Llama 3.2 90B 66%
Perception: Battle
9 models we track have a score here.
- 1 Claude 3.7 Sonnet ↑2 85.7%
- 2 o3 ↓1 85.4%
- 3
Qwen2.5-VL-72B ↑1 82.4%
- 4
InternVL3-78B ↑3 81.8%
- 5 GPT-4.1 81.2%
- 6 Grok 2 Vision (1212) ↑2 81%
- 7
UI-TARS-1.5 ↓5 81%
- 8 Qwen-VL-Max ↓2 80.6%
- 9
Llama 3.2 90B 72.4%
Strategic Reasoning
9 models we track have a score here.
- 1 o3 46.6%
- 2 Claude 3.7 Sonnet ↑1 40.4%
- 3
UI-TARS-1.5 ↓1 39%
- 4 GPT-4.1 ↑1 36%
- 5
Qwen2.5-VL-72B ↓1 32.2%
- 6 Qwen-VL-Max 31.2%
- 7
InternVL3-78B 29.1%
- 8
Llama 3.2 90B ↑1 26.4%
- 9 Grok 2 Vision (1212) ↓1 26.2%
Strategic Reasoning: Hanabi
9 models we track have a score here.
- 1 o3 61.2%
- 2 Claude 3.7 Sonnet ↑1 39%
- 3
Qwen2.5-VL-72B ↑1 26.8%
- 4 Qwen-VL-Max ↑2 26.5%
- 5
UI-TARS-1.5 ↓3 25.5%
- 6
InternVL3-78B ↑1 25.2%
- 7 GPT-4.1 ↓2 23%
- 8
Llama 3.2 90B ↑1 20%
- 9 Grok 2 Vision (1212) ↓1 12.8%
Strategic Reasoning: Overcooked
9 models we track have a score here.
- 1 o3 40.5%
- 2 GPT-4.1 ↑3 27%
- 3
Qwen2.5-VL-72B ↑1 26.5%
- 4 Qwen-VL-Max ↑2 26%
- 5 Claude 3.7 Sonnet ↓2 26%
- 6
InternVL3-78B ↑1 20.5%
- 7
UI-TARS-1.5 ↓5 18.2%
- 8 Grok 2 Vision (1212) 17.2%
- 9
Llama 3.2 90B 16.5%
Strategic Reasoning: KAZ
9 models we track have a score here.
- 1 Qwen-VL-Max ↑5 46%
- 2
Qwen2.5-VL-72B ↑2 39.5%
- 3
UI-TARS-1.5 ↓1 38.5%
- 4 Claude 3.7 Sonnet ↓1 35.8%
- 5 o3 ↓4 30.8%
- 6 GPT-4.1 ↓1 27.8%
- 7 Grok 2 Vision (1212) ↑1 25.5%
- 8
InternVL3-78B ↓1 24%
- 9
Llama 3.2 90B 20.8%
Strategic Reasoning: Board
9 models we track have a score here.
- 1 o3 29%
- 2 Claude 3.7 Sonnet ↑1 24.2%
- 3
Qwen2.5-VL-72B ↑1 23.8%
- 4
UI-TARS-1.5 ↓2 23.2%
- 5 GPT-4.1 22.5%
- 6 Qwen-VL-Max 19.5%
- 7
InternVL3-78B 14%
- 8
Llama 3.2 90B ↑1 11.8%
- 9 Grok 2 Vision (1212) ↓1 10.8%
Strategic Reasoning: Poker
9 models we track have a score here.
- 1 o3 67%
- 2
UI-TARS-1.5 65.8%
- 3 Claude 3.7 Sonnet 65.2%
- 4 Grok 2 Vision (1212) ↑4 59.5%
- 5 GPT-4.1 58.2%
- 6
Llama 3.2 90B ↑3 52.8%
- 7 Qwen-VL-Max ↓1 52.2%
- 8
Qwen2.5-VL-72B ↓4 50.8%
- 9
InternVL3-78B ↓2 45.2%
Strategic Reasoning: Pong
9 models we track have a score here.
- 1 Claude 3.7 Sonnet ↑2 44.8%
- 2 GPT-4.1 ↑3 41.5%
- 3
Llama 3.2 90B ↑6 36.2%
- 4
InternVL3-78B ↑3 34.8%
- 5
UI-TARS-1.5 ↓3 30.8%
- 6
Qwen2.5-VL-72B ↓2 27%
- 7 o3 ↓6 25.8%
- 8 Qwen-VL-Max ↓2 23.5%
- 9 Grok 2 Vision (1212) ↓1 20.8%
Strategic Reasoning: MPE
9 models we track have a score here.
- 1 Claude 3.7 Sonnet ↑2 35.8%
- 2 GPT-4.1 ↑3 35.8%
- 3
Qwen2.5-VL-72B ↑1 34.2%
- 4 o3 ↓3 33.8%
- 5
UI-TARS-1.5 ↓3 33.5%
- 6
Llama 3.2 90B ↑3 32.8%
- 7 Qwen-VL-Max ↓1 31%
- 8
InternVL3-78B ↓1 30.2%
- 9 Grok 2 Vision (1212) ↓1 25.2%
Strategic Reasoning: Dilemma
9 models we track have a score here.
- 1
UI-TARS-1.5 ↑1 63%
- 2 o3 ↓1 59.2%
- 3 Claude 3.7 Sonnet 53.8%
- 4 GPT-4.1 ↑1 49.8%
- 5
InternVL3-78B ↑2 37%
- 6 Grok 2 Vision (1212) ↑2 30.2%
- 7
Qwen2.5-VL-72B ↓3 30%
- 8 Qwen-VL-Max ↓2 26.2%
- 9
Llama 3.2 90B 22.2%
Strategic Reasoning: Hunt
9 models we track have a score here.
- 1 o3 57.5%
- 2 Claude 3.7 Sonnet ↑1 42.5%
- 3
UI-TARS-1.5 ↓1 41.5%
- 4 GPT-4.1 ↑1 36.8%
- 5 Grok 2 Vision (1212) ↑3 31.5%
- 6
InternVL3-78B ↑1 30%
- 7
Qwen2.5-VL-72B ↓3 27.2%
- 8
Llama 3.2 90B ↑1 26.2%
- 9 Qwen-VL-Max ↓3 23.5%
Strategic Reasoning: Battle
9 models we track have a score here.
- 1 o3 61.5%
- 2
UI-TARS-1.5 50%
- 3 Claude 3.7 Sonnet 43.2%
- 4 GPT-4.1 ↑1 38%
- 5 Qwen-VL-Max ↑1 37.2%
- 6
Qwen2.5-VL-72B ↓2 36.8%
- 7
InternVL3-78B 30.2%
- 8 Grok 2 Vision (1212) 29%
- 9
Llama 3.2 90B 25%
Decision-Making
9 models we track have a score here.
- 1 o3 31.4%
- 2 Claude 3.7 Sonnet ↑1 18.1%
- 3
UI-TARS-1.5 ↓1 12.9%
- 4 GPT-4.1 ↑1 4.80%
- 5
Qwen2.5-VL-72B ↓1 3.00%
- 6
InternVL3-78B ↑1 1.20%
- 7 Qwen-VL-Max ↓1 0.20%
- 8
Llama 3.2 90B ↑1 -6.30%
- 9 Grok 2 Vision (1212) ↓1 -6.70%
Decision-Making: Hanabi
9 models we track have a score here.
- 1 o3 55.8%
- 2 Claude 3.7 Sonnet ↑1 6.70%
- 3 Qwen-VL-Max ↑3 1.20%
- 4
Qwen2.5-VL-72B 0.80%
- 5
Llama 3.2 90B ↑4 0%
- 6 Grok 2 Vision (1212) ↑2 0%
- 7
InternVL3-78B 0%
- 8 GPT-4.1 ↓3 0%
- 9
UI-TARS-1.5 ↓7 0%
Decision-Making: Overcooked
9 models we track have a score here.
- 1 o3 15.6%
- 2 Claude 3.7 Sonnet ↑1 10.1%
- 3
UI-TARS-1.5 ↓1 2.00%
- 4
Llama 3.2 90B ↑5 1.50%
- 5 Grok 2 Vision (1212) ↑3 1.50%
- 6
InternVL3-78B ↑1 0%
- 7 Qwen-VL-Max ↓1 -0.50%
- 8
Qwen2.5-VL-72B ↓4 -0.50%
- 9 GPT-4.1 ↓4 -0.50%
Decision-Making: KAZ
9 models we track have a score here.
- 1 o3 29.5%
- 2 Claude 3.7 Sonnet ↑1 5.30%
- 3
Llama 3.2 90B ↑6 0%
- 4 Grok 2 Vision (1212) ↑4 0%
- 5
InternVL3-78B ↑2 -3.50%
- 6
UI-TARS-1.5 ↓4 -4.70%
- 7 Qwen-VL-Max ↓1 -5.30%
- 8
Qwen2.5-VL-72B ↓4 -5.30%
- 9 GPT-4.1 ↓4 -5.30%
Decision-Making: Board
9 models we track have a score here.
- 1 o3 65%
- 2 Claude 3.7 Sonnet ↑1 20%
- 3
UI-TARS-1.5 ↓1 5.00%
- 4 Qwen-VL-Max ↑2 0%
- 5
Llama 3.2 90B ↑4 0%
- 6 Grok 2 Vision (1212) ↑2 0%
- 7
Qwen2.5-VL-72B ↓3 0%
- 8
InternVL3-78B ↓1 0%
- 9 GPT-4.1 ↓4 0%
Decision-Making: Poker
9 models we track have a score here.
- 1 Claude 3.7 Sonnet ↑2 72.9%
- 2 o3 ↓1 61.8%
- 3
UI-TARS-1.5 ↓1 36.4%
- 4
InternVL3-78B ↑3 4.00%
- 5
Qwen2.5-VL-72B ↓1 -3.10%
- 6 GPT-4.1 ↓1 -7.10%
- 7 Grok 2 Vision (1212) ↑1 -11.80%
- 8 Qwen-VL-Max ↓2 -20.50%
- 9
Llama 3.2 90B -39.40%
Decision-Making: Pong
9 models we track have a score here.
- 1 o3 8.60%
- 2 GPT-4.1 ↑3 0.20%
- 3 Grok 2 Vision (1212) ↑5 -0.10%
- 4 Qwen-VL-Max ↑2 -0.30%
- 5 Claude 3.7 Sonnet ↓2 -0.50%
- 6
Qwen2.5-VL-72B ↓2 -0.80%
- 7
Llama 3.2 90B ↑2 -0.90%
- 8
InternVL3-78B ↓1 -0.90%
- 9
UI-TARS-1.5 ↓7 -0.90%
Decision-Making: MPE
9 models we track have a score here.
- 1 Claude 3.7 Sonnet ↑2 39.4%
- 2 o3 ↓1 37.2%
- 3 GPT-4.1 ↑2 31.5%
- 4
Qwen2.5-VL-72B 19.6%
- 5
UI-TARS-1.5 ↓3 19.4%
- 6 Qwen-VL-Max 14.4%
- 7
InternVL3-78B 5.90%
- 8
Llama 3.2 90B ↑1 -29.60%
- 9 Grok 2 Vision (1212) ↓1 -58.20%
Decision-Making: Dilemma
9 models we track have a score here.
- 1
UI-TARS-1.5 ↑1 33.1%
- 2 GPT-4.1 ↑3 17.8%
- 3
InternVL3-78B ↑4 5.00%
- 4 Claude 3.7 Sonnet ↓1 4.60%
- 5 Grok 2 Vision (1212) ↑3 1.10%
- 6
Llama 3.2 90B ↑3 0.40%
- 7
Qwen2.5-VL-72B ↓3 0%
- 8 Qwen-VL-Max ↓2 -0.40%
- 9 o3 ↓8 -0.40%
Decision-Making: Hunt
9 models we track have a score here.
- 1
UI-TARS-1.5 ↑1 24.9%
- 2 o3 ↓1 24%
- 3 Claude 3.7 Sonnet 19.9%
- 4
Qwen2.5-VL-72B 19.6%
- 5 Qwen-VL-Max ↑1 13.2%
- 6 GPT-4.1 ↓1 11.2%
- 7
Llama 3.2 90B ↑2 3.60%
- 8
InternVL3-78B ↓1 -0.20%
- 9 Grok 2 Vision (1212) ↓1 -0.40%
Decision-Making: Battle
9 models we track have a score here.
- 1 o3 16.7%
- 2
UI-TARS-1.5 13.6%
- 3 Claude 3.7 Sonnet 2.50%
- 4
InternVL3-78B ↑3 1.50%
- 5
Llama 3.2 90B ↑4 1.00%
- 6 Grok 2 Vision (1212) ↑2 0.50%
- 7 GPT-4.1 ↓2 0.50%
- 8 Qwen-VL-Max ↓2 -0.50%
- 9
Qwen2.5-VL-72B ↓5 -0.50%
Number one over time
Lab claims count from the day the card came out, independent runs from the model’s release.
Source: Official Visual Strategic Bench leaderboard
Scores over time
All the scores we have, one dot each, so a model tested at several settings appears more than once.
Source: Official Visual Strategic 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 Games and long tasks
- ARC-AGI-3 In use Interactive game environments where agents must explore, infer goals and rules, and act efficiently without instructions. GPT-6.2 Astra 100%
- Orak Fading LLM agents play 12 popular video games across most genres through an MCP interface. Each game scores in-game progress or outcomes, and the board ranks models by their average rank across the games.
Gemini 2.5 Pro 3.90
- MineDojo Fading Agents complete open-ended tasks in Minecraft, such as mining blocks and fighting mobs, from a large simulation suite. Programmatic checks score many tasks.
Seed2.0 Pro 49%
- TextQuests Only in system cards Agents play 25 classic interactive fiction games. The score is harmful actions per 100 steps, where lower is better. The board also reports game progress.
Llama 4 Scout 7.00
- GameWorld Only in system cards Multimodal game agents play 34 browser games across 170 tasks. Outcomes are checked from the game state. GPT-6.1 Terra 76.8%
- Swarm Bench Only in system cards Models orchestrate agent swarms on complex tasks that benefit from coordinated decomposition and parallel execution. Kimi K3 76.3%
Data audit
9 of 10 checks pass, 1 warning. Look first at the Newest score check. The newest came 19 months ago, in April 2025, for o3. 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 (13 entries).
- Pass Models scored 9 models we track, plus 4 other entries on its leaderboard.
- Warning Newest score The newest came 19 months ago, in April 2025, for o3.
- Note Frontier models None of the top 10 has a score. The strongest tested is o3 (#84).
- Pass Headroom The best is 84.9%, under 85%, so it still separates models.
Leaderboard import 3 pass
- Pass Importer We read the board on vs-bench.github.io: 10 rows for models we track and 4 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.