Benchmark Almanac

WorldState-Bench

Stateful Harness + MiniMax H3 92% tops its leaderboard MiniMax H3 32% leads as a model on its own
Not in model cards

Video models generate two clips from the same image and action. A hidden state changes which ending the rule requires. A vision model checks both clips.

The WorldRuler team ran all 50 official pairs with the original judge protocol. The makers also publish results for 11 video models and their own system.

Released
October 2026
Kind
Video world models
Level
Everyday
Scored
Graded by a model
Answers
Open-ended
Tasks
All public
Measures
Capability
Language
English
In and out
Images, Text → Video
Interaction
Single turn
Items from
Written by experts
Refresh
One fixed set
Run by
Anyone
Last model added
July 2025
Scores refreshed
November 20th, 2026
12
Entries on its own leaderboard, 3 models we track

Stateful Harness + MiniMax H3 tops the leaderboard with 92% 2 models share the top score of 32% among models released by July 2025

Best score per entry from the benchmark's own leaderboard, read on November 20th, 2026. Where a model ran on its own with several settings, we keep its best. Agents other companies built around a model get a pale shade of that model's lab color. Gray marks blends and entries not in our model list, such as fine-tunes and older models. None of these count toward leaders. The best model on its own is MiniMax H3, at 32%. Agents built around a model, such as Stateful Harness + MiniMax H3, are their own entries. We don't count them as the model's score. This view shows only models we track, each at its best.

The newest model with a score here came out in July 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 Stateful Harness + MiniMax H3 92%
  2. 2 MiniMax H3 32%
  3. 3 Wan2.2-I2V-A14B 32%
  4. 4 Echo-WM Flash 26%
  5. 5 HunyuanVideo 1.5 20%
  6. 6 LingBot-World v2 16%
  7. 7 AlayaWorld 12%
  8. 8 FantasyWorld 8.00%
  9. 9 LongCat-Video 13.6B 8.00%
  10. 10 InSpatio-World 2.00%
  11. 11 Astra 0%
  12. 12 HY-WorldPlay 0%

Source: Official WorldState-Bench leaderboard

Source: Official WorldState-Bench leaderboard

#EntryScoreSettingDateSource
1 Stateful Harness + MiniMax H3 92% Official leaderboard
2 MiniMax H3 32% Official leaderboard
3 Wan2.2-I2V-A14B 32% Jul 2025 Official leaderboard
4 Echo-WM Flash 26% Official leaderboard
5 HunyuanVideo 1.5 20% Official leaderboard
6 LingBot-World v2 16% Official leaderboard
7 AlayaWorld 12% Official leaderboard
8 FantasyWorld 8.00% Official leaderboard
9 LongCat-Video 13.6B 8.00% Official leaderboard
10 InSpatio-World 2.00% Official leaderboard
11 Astra 0% Official leaderboard
12 HY-WorldPlay 0% Official leaderboard
#ModelScoreSettingDateSource
1 MiniMax H3 32% Official leaderboard
2 Wan2.2-I2V-A14B 32% Jul 2025 Official leaderboard
3 HunyuanVideo 1.5 20% Official 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
0% 20% 40% 60% 80% 100% Near the ceiling (85%) Wan2.2-I2V-A14B 0% 20% 40% 60% 80% 100% Near the ceiling (85%) Wan2.2-I2V-A14B

Source: Official WorldState-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 World models

All 53 in World models
  • AgentWorldBench In use Language world models predict the next environment observation after an agent action, in 7 domains from MCP and search to Android and desktop OS. A judge model scores each prediction against the real observation on five dimensions. GPT-5.4 58.3
  • IntPhys 2 In use Models watch simulated videos and judge whether each event is physically possible, testing object permanence, immutability, continuity and solidity. Scored by accuracy. Muse Spark Open 68.3%
  • QwenWorldBench Fading Models act as world models that simulate agent environments in 7 domains, such as terminal, web, and Android. A rubric judge scores five dimensions. Qwen3.7-Plus 62.1%
  • VPCT Fading Models view an image of a ball, ramps, and three buckets and predict which bucket the ball falls into. Scored by accuracy on 100 problems. Gemini 3 Pro 91%
  • ContPhy Fading Models infer physical properties such as mass and density of fluids and soft bodies from videos and predict their dynamics. Seed2.0 Pro 67.4%
  • RBench Only in system cards 650 image+text prompts for robot video generation (tasks and embodiments). MLLM judge scores task completion, physical plausibility, stability. Wan 2.6 60.7%

Data audit

9 of 10 checks pass, 1 warning. Look first at the Newest score check. The newest came 16 months ago, in July 2025, for Wan2.2-I2V-A14B. 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 (12 entries).
  • Pass Models scored 3 models we track, plus 9 other entries on its leaderboard.
  • Warning Newest score The newest came 16 months ago, in July 2025, for Wan2.2-I2V-A14B.
  • Note Frontier models None of its models are on the capabilities index.
  • Pass Headroom The best is 32%, under 85%, so it still separates models.
Leaderboard import 3 pass
  • Pass Importer We read the board on ivg-yanranzhang.github.io: 4 rows for models we track and 8 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