New: the Timeline is live. Track world model releases, papers, and benchmark updates in real time.
world-models.io
The Knowledge Hub for AI World Models

Compare World Models

Structured side-by-side comparisons of AI world models across architecture, performance, and use cases.

robotics model-based-rl simulation embodied-ai

Quick Answer

Short extractable summary preserved directly in static HTML.

  • The comparison hub helps readers choose between world models, architectures, labs, and research paradigms by exposing verdicts, side-by-side dimensions, FAQs, and primary sources.
  • Use it to get a short answer first, then inspect trade-offs, evidence, and linked model pages.

Editorial Trust Signals

Editorial provenance and refresh policy preserved directly in static HTML.

Published by world-models.io editorial board.

Lead editor Bernard Grenat.

This hub publishes comparison pages designed for direct answers, explicit trade-offs, and source-backed verification.

Each editorial page is assembled from primary sources, normalized into extractable summaries, checked for factual drift, and reviewed before publication or major refreshes. Last reviewed: 2026-06-21.

Pages are refreshed when a new paper, benchmark, release, architecture update, or stronger primary source materially changes the answer a reader or AI system should retrieve.

Each page links back to relevant primary sources and keeps a stable canonical URL so readers can verify claims, trace context, and reference the most up-to-date version. See the editorial policy.

Primary sources onlyLast reviewed date visibleMethodology documentedSource links included

Quick Answer

Static overview of the comparison catalog.

  • The comparison hub helps readers choose between world models, architectures, labs, and research paradigms by exposing verdicts, side-by-side dimensions, FAQs, and primary sources.
  • This static directory exposes 61 comparisons before JavaScript runs, including editorial and programmatic comparison pages.

Editorial Trust Signals

Editorial provenance and refresh policy preserved directly in static HTML.

Published by world-models.io editorial board.

Lead editor Bernard Grenat.

This hub publishes comparison pages designed for direct answers, explicit trade-offs, and source-backed verification.

Each editorial page is assembled from primary sources, normalized into extractable summaries, checked for factual drift, and reviewed before publication or major refreshes. Last reviewed: 2026-06-21.

Pages are refreshed when a new paper, benchmark, release, architecture update, or stronger primary source materially changes the answer a reader or AI system should retrieve.

Each page links back to relevant primary sources and keeps a stable canonical URL so readers can verify claims, trace context, and reference the most up-to-date version. See the editorial policy.

Primary sources onlyLast reviewed date visibleMethodology documentedSource links included

Featured Comparisons

Representative comparisons rendered directly in static HTML.

ComparisonItem AItem BSummary
3D-VLA vs I-JEPA3D-VLAI-JEPATwo approaches to learning representations for embodied intelligence: 3D-VLA combines 3D perception with language-conditioned action planning, while I-JEPA learns abstract visual representations through self-supervised prediction in latent space.
AMI vs Ha World ModelAMIHa World ModelTwo pioneering cognitive-inspired world models: Ha's 2018 World Model introduced the VAE+RNN+Controller architecture, while AMI proposes an autonomous machine intelligence framework inspired by biological cognition.
Copilot4D vs GAIA-1Copilot4DGAIA-1Both target autonomous driving simulation but from different angles: Copilot4D predicts 4D point cloud futures for safety-critical planning, while GAIA-1 generates photorealistic driving video for scenario exploration.
Copilot4D vs GAIA-1Copilot4DGAIA-1Two autonomous driving world models with different approaches: Copilot4D uses discrete tokenization for LiDAR point cloud forecasting, while GAIA-1 generates photorealistic driving videos from multimodal inputs.
DreamerV2 vs DreamerV3DreamerV2DreamerV3The Dreamer lineage's two most impactful iterations: DreamerV2 achieved human-level Atari with discrete representations, while DreamerV3 eliminated hyperparameter tuning entirely with symlog predictions.
DreamerV2 vs PlaNetDreamerV2PlaNetBoth use the RSSM architecture for latent dynamics, but DreamerV2 introduced discrete representations that dramatically improved performance. PlaNet pioneered the approach; DreamerV2 perfected it for Atari-scale environments.
DreamerV3 vs DIAMONDDreamerV3DIAMONDDreamerV3 and DIAMOND are both model-based RL agents that train policies via imagination, but they use fundamentally different dynamics models: RSSM latent dynamics vs. pixel-space diffusion models.
DreamerV3 vs IRISDreamerV3IRISTwo model-based RL agents using fundamentally different world model architectures: DreamerV3's RSSM with actor-critic vs. IRIS's autoregressive Transformer with VQ-VAE tokens.
DreamerV3 vs MuZeroDreamerV3MuZeroBoth are landmark world model systems, but with fundamentally different architectures. DreamerV3 uses latent imagination with actor-critic learning, while MuZero uses abstract learned dynamics with Monte Carlo tree search.
DreamerV3 vs PlaNetDreamerV3PlaNetDreamerV3 represents the evolution of PlaNet's core ideas. Both use the RSSM architecture, but DreamerV3 adds discrete representations, symlog predictions, and fixed hyperparameters to achieve state-of-the-art performance across diverse domains.
DreamerV3 vs PlayWorldDreamerV3PlayWorldTwo green-index leaders for acting under learned dynamics, but with different centers of gravity. DreamerV3 is the canonical imagination-based general RL agent, while PlayWorld is a manipulation-centric robot simulator trained from autonomous play data.
DreamerV3 vs TD-MPC2DreamerV3TD-MPC2Two leading model-based RL agents with different philosophies: DreamerV3 uses imagination-based actor-critic learning, while TD-MPC2 combines temporal-difference learning with model-predictive control for multi-task mastery.
Emu Video vs SoraEmu VideoSoraBoth are frontier video generation models, but with different ambitions: Emu Video focuses on efficient, high-quality short-form generation, while Sora pushes toward long-form, physically coherent world simulation.
GAIA-1 vs Copilot4DGAIA-1 (Wayve)Copilot4D (Waabi)Both are world models designed for autonomous driving, but they operate on different sensor modalities: GAIA-1 generates camera video, while Copilot4D predicts LiDAR point clouds in 4D.
GAIA-1 vs NVIDIA CosmosGAIA-1NVIDIA CosmosBoth are video-based world models for autonomous driving and physical AI, but GAIA-1 is a domain-specific driving world model from Wayve while Cosmos is a general-purpose foundation platform from NVIDIA.
GAIA-1 vs SoraGAIA-1SoraTwo generative world models that approach video generation from different angles: GAIA-1 focuses on autonomous driving simulation, while Sora aims to be a general-purpose visual world simulator.

Comparison Directory

Canonical crawlable directory of comparison pages. Client-side filtering can enhance this same catalog without hiding it from no-JS readers.

ComparisonItem AItem BDimensionsFAQs
3D-VLA vs I-JEPA3D-VLAI-JEPA71
AMI vs Ha World ModelAMIHa World Model61
Copilot4D vs GAIA-1Copilot4DGAIA-171
Copilot4D vs GAIA-1Copilot4DGAIA-161
DreamerV2 vs DreamerV3DreamerV2DreamerV362
DreamerV2 vs PlaNetDreamerV2PlaNet72
DreamerV3 vs DIAMONDDreamerV3DIAMOND72
DreamerV3 vs IRISDreamerV3IRIS71
DreamerV3 vs MuZeroDreamerV3MuZero71
DreamerV3 vs PlaNetDreamerV3PlaNet71
DreamerV3 vs PlayWorldDreamerV3PlayWorld72
DreamerV3 vs TD-MPC2DreamerV3TD-MPC272
Emu Video vs SoraEmu VideoSora71
GAIA-1 vs Copilot4DGAIA-1 (Wayve)Copilot4D (Waabi)72
GAIA-1 vs NVIDIA CosmosGAIA-1NVIDIA Cosmos71
GAIA-1 vs SoraGAIA-1Sora71
GameNGen vs DIAMONDGameNGenDIAMOND71
Genie 2 vs UniSimGenie 2UniSim72
Genie 3 vs Genie 2Genie 3Genie 272
Genie 3 vs NVIDIA CosmosGenie 3NVIDIA Cosmos72
Genie 3 vs V-JEPA 2Genie 3V-JEPA 272
Genie vs Genie 2Genie (v1)Genie 271
Ha & Schmidhuber World Model vs DreamerV3Ha & Schmidhuber World ModelDreamerV371
I-JEPA vs MAE (Masked Autoencoders)I-JEPA (Meta FAIR)MAE (Meta / He et al.)71
IRIS vs DIAMONDIRISDIAMOND61
IRIS vs DreamerV3IRISDreamerV371
LeWorldModel vs DreamerV3LeWorldModelDreamerV372
LWM vs V-JEPALarge World Model (LWM)V-JEPA61
MILE vs GAIA-1MILEGAIA-161
Model-Based RL vs Model-Free RLModel-Based RLModel-Free RL72
MuZero vs DreamerV3MuZeroDreamerV382
MuZero vs TD-MPC2MuZeroTD-MPC272
NVIDIA Cosmos vs DreamerV3NVIDIA CosmosDreamerV371
NVIDIA Cosmos vs Genie 2NVIDIA CosmosGenie 272
NVIDIA Cosmos vs V-JEPA 2NVIDIA CosmosV-JEPA 272
OASIS vs DIAMONDOASISDIAMOND72
OASIS vs GameNGenOASIS (Decart)GameNGen (Google Research)72
OASIS vs PandoraOASISPandora61
Pandora vs Genie 2PandoraGenie 272
Pandora vs OASISPandoraOASIS61
PixVerse R1 vs SoraPixVerse R1Sora72
PlayWorld vs TD-MPC2PlayWorldTD-MPC272
PlayWorld vs V-JEPA 2PlayWorldV-JEPA 272
Predictron vs MuZeroPredictronMuZero71
Predictron vs MuZeroPredictronMuZero61
RT-2 vs 3D-VLART-23D-VLA61
Sora vs Emu VideoSoraEmu Video71
Sora vs Gen-3 AlphaSoraGen-3 Alpha72
Sora vs Genie 2Sora (OpenAI)Genie 2 (DeepMind)72
Sora vs Genie 3SoraGenie 382
Sora vs NVIDIA CosmosSora (OpenAI)NVIDIA Cosmos72
Stable Video Diffusion vs Emu VideoStable Video DiffusionEmu Video61
STEVE-1 vs DreamerV3STEVE-1DreamerV361
UniSim vs Genie 2UniSimGenie 271
UniSim vs Genie 2UniSimGenie 261
V-JEPA 2 vs I-JEPAV-JEPA 2I-JEPA82
V-JEPA 2 vs V-JEPAV-JEPA 2V-JEPA61
V-JEPA vs I-JEPAV-JEPAI-JEPA72
V-JEPA vs NVIDIA CosmosV-JEPANVIDIA Cosmos71
V-JEPA vs Video Generation ModelsV-JEPA (Meta)Video Generation Models (Sora, Cosmos)72
World Models vs LLMsWorld ModelsLarge Language Models71

External Sources

Primary references and official sources surfaced directly in static HTML for crawlers and no-JS readers.