Hedra launches Model Factory to build and refine AI models for robotics
Also covered by 1 source
What happened
Hedra has launched Model Factory, a platform that helps organizations in robotics, manufacturing, biology, and defense build and refine AI models tailored to their specific needs. The service integrates GPU computing, model training, deployment, and feedback into a single workflow. Customers can collaborate with Hedra researchers and AI agents to iteratively improve their models using their own data and simulation environments. Model Factory became available on November 10, offering deployment on-premises or through partner data centers.
Key details
- Model Factory supports supervised fine-tuning, LoRA adapters, and reinforcement learning for model training.
- A reported example showed a sixfold speedup in 2K image editing, reducing time from 22.9 seconds to 3.3 seconds on four AMD Instinct MI355X GPUs.
- Retraining the H3 model increased its Physics-IQ benchmark score from 30.3 to 34.9.
- Teams can use GPU sandboxes without long-term commitments and bring their own models or data.
- Deployment options include on-premises infrastructure or partner data centers.
Why it matters
Teams in robotics, manufacturing, biology, and defense can now develop and deploy AI models that are specifically adapted to their real-world environments and hardware. This reduces the time and technical barriers to achieving reliable AI performance in production settings, allowing organizations to iterate quickly and meet their own evaluation standards. The ability to use flexible GPU resources and collaborate directly with researchers gives companies more control over their AI development process and deployment options.