Benchmark Almanac

Visual Strategic Bench

o3 84.9% leads since June 2025
Not in model cards

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
In and out
Images, Text → Text, Actions
Interaction
Multi-agent
Environment
A simulator
Refresh
Live
Run by
Anyone
Strongest model tested
o3 #84 on the capability index
Last model added
April 2025
Scores refreshed
November 20th, 2026
13
Entries on its own leaderboard, 9 models we track

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:

  1. 1 o3 84.9%
  2. 2 gemini-2.5-pro (snapshot unspecified) Thinking off 84.5%
  3. 3 UI-TARS-1.5 81.5%
  4. 4 Claude 3.7 Sonnet 81.1%
  5. 5 Qwen2.5-VL-72B 80.3%
  6. 6 GPT-4.1 80.3%
  7. 7 Qwen-VL-Max 80.2%
  8. 8 doubao-1-5-vision-pro 77.6%
  9. 9 doubao-1-5-thinking-pro 74.7%
  10. 10 qvq-max 74.5%
  11. 11 InternVL3-78B 74.1%
  12. 12 Grok 2 Vision (1212) 70.2%
  13. 13 Llama 3.2 90B 67.8%

Source: Official Visual Strategic Bench leaderboard

#EntryScoreSettingDateSource
1 o3 84.9% Apr 2025 Official leaderboard
2 gemini-2.5-pro (snapshot unspecified) 84.5% Thinking off Official leaderboard
3 UI-TARS-1.5 81.5% Apr 2025 Official leaderboard
4 Claude 3.7 Sonnet 81.1% Feb 2025 Official leaderboard
5 Qwen2.5-VL-72B 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 InternVL3-78B 74.1% Apr 2025 Official leaderboard
12 Grok 2 Vision (1212) 70.2% Dec 2024 Official leaderboard
13 Llama 3.2 90B 67.8% Sep 2024 Official leaderboard
#ModelScoreSettingDateSource
1 o3 84.9% Apr 2025 Official leaderboard
2 UI-TARS-1.5 81.5% Apr 2025 Official leaderboard
3 Claude 3.7 Sonnet 81.1% Feb 2025 Official leaderboard
4 Qwen2.5-VL-72B 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 InternVL3-78B 74.1% Apr 2025 Official leaderboard
8 Grok 2 Vision (1212) 70.2% Dec 2024 Official leaderboard
9 Llama 3.2 90B 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. 1 o3 84.9%
  2. 2 UI-TARS-1.5 81.5%
  3. 3 Claude 3.7 Sonnet 81.1%
  4. 4 Qwen2.5-VL-72B 80.3%
  5. 5 GPT-4.1 80.3%
  6. 6 Qwen-VL-Max 80.2%
  7. 7 InternVL3-78B 74.1%
  8. 8 Grok 2 Vision (1212) 70.2%
  9. 9 Llama 3.2 90B 67.8%

Number one over time

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

OpenAI
Own leaderboard Not tested since Jun 2025
o3 84.9%

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.

Plot with other benchmarks
On the benchmark’s own leaderboard Best so far
0% 20% 40% 60% 80% 100% 2024 2025 Near the ceiling (85%) o3 Qwen2.5-VL-72B Qwen-VL-Max 0% 20% 40% 60% 80% 100% 2024 2025 Near the ceiling (85%) o3 Qwen-VL-Max

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

All 261 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.

How we check