Skip to main content

LTX-2.5 Open World Model Launches: Free for SMBs, Native ComfyUI Support

LTX, a company that spun off from Lightricks, has officially released its open-weights video and world model, LTX-2.5. The model is now available on Hugging Face, and it comes with a strategic partnership with ComfyUI for native integration. Plus, they're offering managed API services for those who want a plug-and-play experience.

On the business side, LTX is taking a tiered approach. If your organization's annual recurring revenue (ARR) is under $10 million, you can use it for free. Larger enterprises will need to negotiate a license. So far, the model series has racked up over 33 million downloads—a clear sign of its popularity.

Image

What's new in LTX-2.5? The generation pipeline has been deeply reconfigured. Key upgrades include a new diffusion video decoder that boosts detail and image quality, native support for multi-camera single-pass rendering, integration of a custom Gemma4 language backbone, and specialized pre-trained checkpoints for physical AI and robotics. Thanks to NVIDIA optimization, the efficient distillation model runs cost-effectively on RTX GPUs and even Macs. Official data shows the generation cost is about one-eighth that of similar models, and rendering is seven times faster.

Performance-wise, in self-hosted tests using two top-tier NVIDIA GB200 chips, LTX-2.5 generated a 10-second 720p video in just 6.8 seconds—that's faster than real-time playback. In quality blind tests, it won 67% of the time. LTX CEO Zeev Farbman says betting on open-source weights is about attracting developers through openness. By taking the path of "local verification first, then commercial licensing," they're targeting diverse enterprise applications like virtual production, real-time robotics, and edge computing.

Key Points

  • Open Weights: LTX-2.5 is available on Hugging Face with open weights.
  • ComfyUI Integration: Native integration with ComfyUI for seamless workflow.
  • Free for SMBs: Organizations with ARR under $10M can use it for free.
  • Performance: Generates 10s 720p video in 6.8s, 7x faster rendering, 1/8 cost.
  • Target Applications: Virtual production, robotics, edge computing, and more.