1X Technologies Launches AI World Model and Home Robot NEO: Presale Begins, Delivery in the US in 2026

Norwegian robotics company 1X Technologies is pushing the boundaries of humanoid robot technology once again. The company announced the launch of the brand new 1X World Model (1XWM) and simultaneously opened pre-sales for its home robot NEO, which is an artificial intelligence model capable of “pre-experiencing reality” in digital environments. This technology is specifically designed for the home robot NEO, allowing it to learn how the world operates through simulation and thus learn to complete tasks in the real world more quickly. This advancement means that robots no longer need to constantly try and error in the real world, but can instead simulate the outcomes of actions and predict environmental changes “in their minds”, significantly shortening the learning cycle.

NEO open pre-order: $499 monthly subscription or $20,000 early bird ownership plan

According to the latest information from the official website, NEO will provide two options:

Standard subscription plan: $499 per month, including the productivity suite and standard shipping;

Early ownership plan: one-time payment of $20,000, offering a three-year warranty, premium support, and priority shipping.

Users can immediately pay a refundable deposit of 200 USD to make a reservation, with deliveries expected to start in 2026 in the United States. NEO will offer three color options, and the overall design continues the 1X minimalist humanoid style, featuring a soft fabric-covered exterior that blends technology with home aesthetics.

From Simulation to Reality: NEO's “World Model” Training Method

Traditional robotic learning relies on a large amount of physical operation and real testing, which is time-consuming and costly. 1XWM changes this. The model can simulate every step of the robot's task completion from images and robotic perception data, and predict how the world will change next. In other words, NEO can not only “see” its actions but also “imagine” the outcomes. For example, when NEO tries to take a bowl out of a cabinet, 1XWM predicts the displacement of the bowl, the change in its center of gravity, and whether the task is ultimately successful after the arm's movement. This allows NEO to undergo thousands of virtual training sessions before performing real operations.

Multi-task learning: making NEO more flexible and smarter

One of the other features of 1XWM is “multi-task training.” The model learns simultaneously from hundreds of different scenarios, including cooking, organizing, and household chores. Research shows that diverse training data allows the model to better understand physical rules and make reasonable responses in unseen situations. This enables NEO to adapt more quickly to the ever-changing scenarios in a household: whether it's rearranging furniture or needing to use rare kitchen utensils, it can respond flexibly.

Simulation Assessment Mechanism: Reduce Physical Testing Costs

Another significant use of 1XWM is “virtual evaluation.” In the past, to test which control policy performed better, real robots had to conduct multiple experiments. Now, through 1XWM, engineers can first predict the task success rate in a simulated world and select the most promising strategies. Experiments indicate that even if the model's prediction accuracy reaches only 70%, it can still achieve a judgment accuracy of up to 90% in strategy comparisons, significantly enhancing development efficiency.

Despite the impressive results, 1X also admits that the model still faces challenges, such as limited simulation accuracy on objects that were not present during training (like new plants or household appliances). The company stated that the next phase will focus on improving the model's understanding of “long-tail objects” and “full-body actions” so that NEO can execute complex tasks more naturally in a home environment.

This article announces that 1X Technologies has launched an AI world model and home robot NEO: Pre-sales have started, with the earliest delivery in the U.S. in 2026, first appearing in Chain News ABMedia.