AI News·19 Nov 2026·via Digg·Covered by 4 sources

    Higgsfield says GPU Harness brings GPU rentals to Claude and ChatGPT

    Higgsfield has launched GPU Harness, a service that enables AI agents to operate within professional cloud applications from various devices. The company states GPU Harness supports rendering, running open-weight models, and training LoRAs through chatbots. Matt Shumer demonstrated using the service with eight NVIDIA L40S GPUs from an iPhone-sized control desk.

    Higgsfield's new GPU Harness service aims to allow multiple AI agents to function concurrently within professional cloud applications, accessible from any device. The company specifies that the service facilitates tasks such as rendering, executing open-weight AI models, and training Low-Rank Adaptations (LoRAs) directly through chatbot interfaces.

    Matt Shumer, a representative for Higgsfield, demonstrated the service's capabilities. He reported controlling eight NVIDIA L40S GPUs via GPU Harness from his phone, using an iPhone-sized control desk. During this demonstration, Shumer stated that Fable 5.2 was used to construct a replica of Shibuya, a process he claimed took 12 hours to build and an additional two hours to render.

    This reported workflow suggests a shift in how resource-intensive creative and AI tasks might be managed. Instead of requiring powerful local hardware for applications like 3D modeling and rendering, GPU Harness positions cloud-based GPUs as remotely accessible resources. The control interface, as described by Shumer, allows for the orchestration of these high-performance operations from a mobile device.

    While the company's headline mentions bringing GPU rentals to Claude and ChatGPT, the specific details provided in the article focus on the service's support for rendering and training models. The demonstration by Shumer illustrates the potential for remote control over significant computational power for creative tasks, rather than elaborating on the direct integration or user experience within those specific chatbot platforms for GPU access.

    Share this article

    Original reporting by DiggWe don't republish, read the full story →

    Related reading

    6 stories