--- title: "PlayWorld vs TD-MPC2 | Green-Index Robot World Models." canonical: "https://world-models.io/en/compare/playworld-vs-td-mpc2/" description: "Comparing PlayWorld and TD-MPC2: manipulation-focused world simulation versus multi-task model-predictive control." author: "Bernard Grenat" generated: "2026-10-30" --- New: the Timeline is live. Track world model releases, papers, and benchmark updates in real time. The Knowledge Hub for AI World Models # PlayWorld vs TD-MPC2 Two green-index models for robot decision-making, but with very different operating modes. PlayWorld learns a manipulation-focused world simulator from autonomous play, while TD-MPC2 combines latent dynamics with model-predictive control across a wide multi-task control benchmark suite. robotics model-based-rl simulation embodied-ai ## What is a world model? A world model is an AI system that learns a predictive representation of how an environment changes over time. It can use observations, actions, or language to estimate future states and possible outcomes before an agent acts. World models support planning in model-based reinforcement learning, robotics, embodied AI, autonomous driving, and interactive simulation. This site catalogs these systems with model records, research papers, benchmark context, and source links. Start with the [world models database](https://world-models.io/en/world-models-database/), compare systems on the [leaderboard](https://world-models.io/en/leaderboard/), or read the [Performance Index methodology](https://world-models.io/en/methodology/performance-index/). ## Comparison Overview Main comparison summary preserved directly in static HTML. Two green-index models for robot decision-making, but with very different operating modes. PlayWorld learns a manipulation-focused world simulator from autonomous play, while TD-MPC2 combines latent dynamics with model-predictive control across a wide multi-task control benchmark suite. ## Verdict Primary editorial conclusion preserved for non-JS crawlers and readers. TD-MPC2 is the safer choice if you want the most established green-index control agent with broad task coverage and mature benchmark evidence. PlayWorld is the more specialized bet if your goal is contact-rich manipulation and learned simulation for policy evaluation. Choose TD-MPC2 for breadth and control maturity; choose PlayWorld for robotics world-simulation depth in manipulation. ## Key Differences Extractable difference list generated from the comparison table. - **Primary Domain:** PlayWorld - Robot manipulation; TD-MPC2 - Multi-task control. - **Learning Signal:** PlayWorld - Autonomous robot play + action-conditioned video prediction; TD-MPC2 - TD learning + latent dynamics + MPC. - **Key Strength:** PlayWorld - Contact-rich policy evaluation; TD-MPC2 - Broad benchmark coverage across 104 tasks. - **Planning Style:** PlayWorld - Evaluate and refine behavior in learned rollouts; TD-MPC2 - Online model-predictive control. - **Evidence Style:** PlayWorld - Robot improvement gains from learned simulation; TD-MPC2 - Standardized control benchmark leadership. ## When To Use Each Static decision guidance for no-JS readers. ### Choose PlayWorld when... Choose PlayWorld when your objective is contact-rich policy evaluation; you are working as robot manipulation; you can validate the choice against robot improvement gains from learned simulation. ### Choose TD-MPC2 when... Choose TD-MPC2 when your objective is broad benchmark coverage across 104 tasks; you are working as multi-task control; you can validate the choice against standardized control benchmark leadership. ## Comparison Table TD-MPC2 is the safer choice if you want the most established green-index control agent with broad task coverage and mature benchmark evidence. PlayWorld is the more specialized bet if your goal is contact-rich manipulation and learned simulation for policy evaluation. Choose TD-MPC2 for breadth and control maturity; choose PlayWorld for robotics world-simulation depth in manipulation. | Dimension | PlayWorld | TD-MPC2 | | --- | --- | --- | | Primary Domain | Robot manipulation | Multi-task control | | Learning Signal | Autonomous robot play + action-conditioned video prediction | TD learning + latent dynamics + MPC | | Key Strength | Contact-rich policy evaluation | Broad benchmark coverage across 104 tasks | | Planning Style | Evaluate and refine behavior in learned rollouts | Online model-predictive control | | Evidence Style | Robot improvement gains from learned simulation | Standardized control benchmark leadership | | Robot Focus | Manipulation-heavy settings | General control and manipulation mix | | Year | 2026 | 2024 | ## Performance Index Snapshot High-level scoring context for the models referenced in this comparison. | Model | Category | Index v1.1 | Confidence | | --- | --- | --- | --- | | PlayWorld | Generative World Model | 80/100 | medium | | TD-MPC2 | Model-Based RL | 80/100 | high | | DreamerV3 | Model-Based RL | 88/100 | high | | V-JEPA 2 | Self-Supervised World Model | 87/100 | medium | ## Frequently Asked Questions FAQ answers rendered directly into static HTML for extractable responses. ### Which is better for manipulation research? PlayWorld, because its evidence and design are centered on contact-rich manipulation and policy improvement in learned simulation. ### Which is more general across control tasks? TD-MPC2. Its headline contribution is scaling one model-based agent across a very large multi-task benchmark set. ## Quick Answer Short extractable summary preserved directly in static HTML. - PlayWorld vs TD-MPC2: this page compares where each system is stronger instead of forcing a universal winner. - Use the verdict for the short answer, then validate the trade-offs in the table, evidence sources, and benchmark context. - Related models and source links help connect this comparison to the broader world models landscape. ## 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 comparison page publishes a direct answer, explicit trade-offs, and source-backed evidence that can be validated against primary materials. 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: 2026-09-27. 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.](https://world-models.io/en/editorial-policy/) Primary sources only Last reviewed date visible Methodology documented Source links included Correction channel published Review date visible ## External Sources Primary papers and official sources for the models discussed on this comparison page. - [PlayWorld paper](https://arxiv.org/abs/2603.09030) Princeton University primary research source - [PlayWorld official page](https://robot-playworld.github.io/) Princeton University official page - [PlayWorld cited source](https://arxiv.org/abs/1803.10122) Yin et al., 2026. PlayWorld: Learning Robot World Models from Autonomous Play. arXiv:2603.09030. - [TD-MPC2 paper](https://arxiv.org/abs/2310.16828) MIT / Meta primary research source