Rather

Loading benchmark results…

World model benchmarks

WBench

Multi-turn interactive world-model evaluation across quality, setting, interaction, consistency and physical behavior.

Published resultsBenchmark website

Visual quality Physics & prediction Consistency Control & interaction

Can it follow a sequence of turns?

WBenchRather-generated AI video · Preview · Unscored

Wan 2.2 TI2V 5B

LTX-Video 0.9.5 2B

0:00 / 0:20

About this comparison

What to look for Follow forward, right, right, left. Watch the dog, buildings and road as each model continues from its own last frame.

Starting frame ↗

The task & prompt

Continue the same scene through all 4 turns.

A modern city street in clear daylight. A broad asphalt road with cracked pavement, lined on both sides with storefronts including Cafe Luna, a bookstore, and City Deli with colorful awnings. Parked cars — including a blue sedan — sit along the curbs. Traffic lights, power lines, and mid-rise brick and concrete buildings frame the view. A signalized intersection lies farther ahead with light urban traffic. Occasionally a distant vehicle passes through the intersection. A yellow school bus drives past on the road from behind, overtaking the viewpoint. A medium-sized Shiba Inu with tan-and-cream fur, a curled tail, and an orange harness. Trots forward with a light bouncing gait, paws lifting quickly off the ground. Third-person view from behind and slightly above the Shiba Inu, following the dog as the central subject. Requested interaction: Move forward.

WBench case 1 / four-turn text-conditioned preview

SpeedShorter clips hold their last frame.

Sources & generation details

Generated by Rather. An official benchmark input run with preview settings. No benchmark scores have been computed for these clips.

Case and seed 42 were fixed before generation. All 2 planned model outputs are shown; no best-of selection.

These exact checkpoints were run by Rather. They are not linked to similarly named source-reported leaderboard entries. Full prompts, model revisions, settings and output hashes are in each run record.

768 x 448 visual preview; not a full benchmark evaluation. One preselected seed per model; no output selection or prompt optimization. Rather text-control prompts translate navigation keys; not upstream prompt parity or native action control. Each turn uses 121 frames at 24 fps. Boundary duplicate frames are omitted when joining turns. Conditioning images are resized to the full output frame without cropping. Generated MP4 streams are remuxed for playback; VP9 WebM copies are encoded at CRF 24 for browser compatibility with no frame interpolation. Original outputs remain linked. Reference footage, where provided, is resized to 768 pixels wide; source hashes and attribution are preserved. Input licenses are listed here; model licenses are linked in each run record.

WBench authors / Meituan LongCat and Fudan University. Generation by Rather.

Wan 2.2 TI2V 5B · Seed 42 · 768 × 448 · 121 frames per turn · 24 fps · 40 steps · Guidance 5. Full run record & prompts (opens in a new tab)

LTX-Video 0.9.5 2B · Seed 42 · 768 × 448 · 121 frames per turn · 24 fps · 40 steps · Guidance 5. Full run record & prompts (opens in a new tab)

Benchmark source MIT license Wan 2.2 TI2V 5B original clip LTX-Video 0.9.5 2B original clip

Source revision: bb6dee5a674cc5cb3019f9c79d4871899fd9e04c

Link to this example ↗

44 reported results

points / 100 · Higher is better · Published by WBench

WBench Navigation, sorted by Average
Model / variant	points / 100	points / 100	points / 100	points / 100	points / 100	points / 100	Evidence
Alaya-EVOKE-Turbo3-stepAlaya Lab	82.00	81.90	82.10	83.90	88.10	74.00	View source ↗Developer evaluated
XGEN-JingbidirectionXGEN Labs	81.90	82.40	90.30	75.60	88.10	73.30	View source ↗Developer evaluated
VWA-WorldX-Era AI	81.70	82.10	83.80	85.90	87.70	68.90	View source ↗Mixed evaluation
EnumabidirectionByteDance	81.70	83.60	79.90	83.20	88.60	73.30	View source ↗Developer evaluated
JoyAI-Echo-1.5 (WM)bidirectionJD Future Academy	81.60	81.50	79.40	86.60	89.80	70.60	View source ↗Evaluator unspecified
gd_worldgd_cv	81.20	82.20	79.50	85.30	89.50	69.60	View source ↗Organizer evaluated
XGEN-Jing4-step ARXGEN Labs	81.00	81.30	91.60	78.70	83.70	69.80	View source ↗Developer evaluated
Zing-0.54-step / 5B / realtimeSeedLeap.ai	81.00	80.60	77.80	84.20	88.50	73.80	View source ↗Evaluator unspecified
JoyAI-Echo-1.5 (WM)flash / 4-stepJD Future Academy	81.00	81.10	77.50	87.90	88.30	70.10	View source ↗Evaluator unspecified
HiDream-O1-WorldHiDream.ai	80.90	81.00	82.20	80.00	88.00	73.30	View source ↗Evaluator unspecified
Alaya-EVOKE3-stepAlaya Lab	80.80	82.80	83.80	78.60	86.90	72.10	View source ↗Evaluator unspecified
HelixWorld4-step / ARNoiz Ai	79.90	79.20	76.50	86.40	86.40	70.90	View source ↗Developer evaluated
EnumaAR / 4-step / flashByteDance	79.70	82.80	73.80	83.40	87.80	70.60	View source ↗Developer evaluated
LingBot-World v2fastAnt Group	79.40	81.80	76.80	82.80	86.50	69.10	View source ↗Evaluator unspecified
Kling 3.0Kling AI	79.00	81.40	91.00	69.40	83.70	69.30	View source ↗Evaluator unspecified
LingBot-Worldbase-cameraAnt Group	78.50	78.90	72.60	80.10	89.90	71.20	View source ↗Evaluator unspecified
Wan 2.7Alibaba	78.10	81.50	91.40	64.40	81.60	71.80	View source ↗Evaluator unspecified
HY-World 1.5ar-distillTencent	78.10	78.10	72.20	86.80	86.90	66.30	View source ↗Evaluator unspecified
HY-Video 1.5Tencent	77.90	77.60	85.60	71.40	87.40	67.40	View source ↗Evaluator unspecified
LingBot-WorldfastAnt Group	77.40	79.40	77.90	79.20	84.90	65.70	View source ↗Evaluator unspecified
Happy OysterAlibaba	76.80	77.30	74.20	84.90	84.30	63.50	View source ↗Evaluator unspecified
Lyra 2.04-step ARNVIDIA	76.40	77.10	73.20	85.60	79.30	66.70	View source ↗Evaluator unspecified
AlayaWorldAlaya Lab	76.30	79.30	69.70	80.00	89.50	63.10	View source ↗Evaluator unspecified
Seedance 1.5ByteDance	76.20	82.10	82.90	66.30	81.30	68.40	View source ↗Evaluator unspecified
SANA-WM4-step ARNVIDIA	76.00	79.30	76.10	82.20	80.70	61.90	View source ↗Evaluator unspecified
Cosmos3-SuperNVIDIA	76.00	76.40	91.60	58.00	85.00	69.20	View source ↗Evaluator unspecified
DreamX-World5B ARAmap	75.00	77.50	80.80	78.60	74.90	63.30	View source ↗Evaluator unspecified
ABot-WorldAmap	74.70	76.80	71.40	84.00	79.50	61.70	View source ↗Evaluator unspecified
Cosmos 2.5NVIDIA	74.60	72.90	83.30	63.10	86.50	67.40	View source ↗Evaluator unspecified
HeliosDistilledPKU-YuanGroup	74.30	73.80	75.40	64.40	81.50	76.20	View source ↗Evaluator unspecified
Cosmos3-NanoNVIDIA	74.20	77.40	87.30	59.10	83.70	63.60	View source ↗Evaluator unspecified
LTX 2.3Lightricks	74.20	77.10	85.20	66.40	77.20	64.90	View source ↗Evaluator unspecified
InSpatio-WorldInSpatio	73.90	71.50	71.40	73.20	88.40	65.20	View source ↗Evaluator unspecified
Genie 3Google	73.90	75.20	72.50	73.40	82.60	65.70	View source ↗Evaluator unspecified
Fantasy-WorldAmap	73.80	72.40	71.30	71.90	86.40	66.80	View source ↗Evaluator unspecified
Astronex-World5BAstronex Robotics	73.50	78.20	73.50	63.40	83.60	68.60	View source ↗Evaluator unspecified
YUME 1.5Shanghai AI Lab	73.30	77.60	72.40	71.40	80.10	65.20	View source ↗Evaluator unspecified
LongCat-VideoMeituan	73.20	75.40	72.30	62.10	87.10	68.90	View source ↗Evaluator unspecified
Infinite-WorldMeituan	72.80	77.00	69.30	75.40	80.00	62.10	View source ↗Evaluator unspecified
MatrixGame3Skywork	71.30	75.50	63.60	83.60	74.50	59.30	View source ↗Evaluator unspecified
Kairos 3.0SenseTime	70.30	74.00	70.30	64.10	82.60	60.40	View source ↗Evaluator unspecified
MatrixGame2Skywork	68.70	73.80	67.10	80.30	65.10	57.20	View source ↗Evaluator unspecified
HY-GameCraftTencent	68.20	73.00	66.60	66.30	72.60	62.40	View source ↗Evaluator unspecified
AstraTsinghua	63.70	67.10	59.60	66.90	73.30	51.40	View source ↗Evaluator unspecified

Official navigation leaderboard; 158 cases. Source-reported configurations are preserved. Checkpoint and per-case settings are not fully pinned. A reported score is not an independent verification by Rather.