Anthropic says its Claude Mythos model cut AMD GPU onboarding to about three days
If the claim holds, the cost of moving AI workloads off Nvidia hardware could fall sharply.
Anthropic told potential IPO investors that its Claude Mythos Preview model handled the software work needed to run one of its models on AMD MI355X GPUs in roughly three days, with almost no human involvement. The company says the same onboarding used to take six to 12 months with teams of dozens of engineers. The three-day figure covers prompt processing for a single model, and Mythos wrote or ported nearly all of the low-level kernels. For developers, the result suggests that the effort of moving AI workloads off Nvidia's CUDA ecosystem may be shrinking, though AMD's own earlier claims about AI-written GPU code came from AMD rather than from Anthropic. Anthropic's prospectus also states it can onboard new chip platforms in a fraction of the previous time, without giving a figure.
Why it matters
Anthropic is describing a shift in how fast a major AI lab can adopt new chips, which matters to anyone following the competition between Nvidia and AMD in AI hardware. Porting a model to a new accelerator has long been a slow, expensive engineering effort, and Anthropic says an AI model now handled most of that low-level work with little human help. The figures come from the company itself in its IPO materials and have not been independently verified, so the claim is best read as a signal of where the work is heading rather than settled evidence. If the speed-up holds beyond a single model, it weakens one of the main barriers keeping AI workloads on Nvidia.
Matters or hype?
Does this story matter, or is it hype? Vote before you see what everyone else said.