# LTX World Model Update Runs Faster at Lower Cost > Source: > Published: 2026-08-11 20:15:28+00:00 LTX today released an update to its open-weight world model that provides the same artificial intelligence (AI) capabilities running faster on inference engines at a lower total cost. [Version 2.5 of the LTX world model](https://ltx.io/newsroom/introducing-ltx-2-5) provides organizations with a proprietary AI platform that they can tune as they best see fit using open weights, says Shani Mandel-Laufer, chief business officer for LTX. Additionally, LTX 2.5 includes an updated diffusion video decoder that reduces visual artifacts in high motion while keeping compression rates high. There is also now a native multishot generation capability that renders a full sequence as one output, holding character, scene, and voice across cuts, while a custom Gemma 4 language backbone and dedicated prompt enhancer read and understand complex, multi-subject prompts more accurately. Finally, there is a pretrained checkpoint that makes it possible to fine-tune on their own domain data, so builders can adapt the model to a world beyond use cases that require cinematic video. World models are starting to gain more traction because they simulate behavior in a way that enables end users to interact with a digital twin of a physical environment. Following an interaction, the world model is trained to predict the next moment rather than, as in the case of a generative AI model, the next piece of text. As such, use cases for world models span everything from film and advertising to gaming, simulation, robotics and other forms of physical AI. Many of the use cases, however, will require an AI model that runs locally simply because latency requirements will not allow an application to access an AI model over the Internet, says Shani Mandel-Laufer, chief business officer for LTX. Many organizations are also disinclined to share potentially sensitive data with a provider of a world model that runs in the cloud, she adds. Designed to run on everything from a server configured with graphics processor units (GPUs) to Macintoshes, the LTX world model is also free to use for any organization with less than $10 million in annual recurring revenue. LTX also claims its world model is the most widely deployed to date, with more than 33 million downloads. It’s not clear to what degree organizations will be embracing world models but it’s clear there will be a lot more diversity in terms of what class of AI model to employ for different use cases. There is no such thing as a one-size-fits-all model and most of those models will need to be fine-tuned for specific use cases using open weights, notes Mandel-Laufer. “You can fine tune an open weight world model on your own,” she says. Ultimately, each organization will need to determine for itself how best to employ a world model. The challenge, as always, is finding and retaining the AI expertise that is needed. The percentage of data scientists, engineers and application developers that have world model expertise, after all, is substantially smaller. Ultimately, however, it’s now more a question of when and how world models will be employed rather than if.