Anthropic says its Claude Mythos Preview model ported AI kernels to AMD GPUs in three days
Anthropic told IPO investors that its Claude Mythos Preview model wrote or ported nearly all the low-level kernel code needed to run an Anthropic AI model on AMD's MI355X GPUs in about three days with minimal human help. Co-founder Tom Brown said the same task previously took six to 12 months using teams of dozens of engineers. Anthropic already runs its models across Nvidia, AMD, Google TPUs and AWS Trainium chips, and has a deal with AMD for up to 2 gigawatts of capacity.
GoKawiil's interpretation of the reporting above, not reported fact.
If accurate and generalizable, this could erode Nvidia's long-standing CUDA advantage, since writing hardware-specific kernel code has been one of the main costs that locks AI developers into Nvidia's ecosystem. Faster porting could turn chip choice into more of a simple purchasing decision rather than an engineering commitment, strengthening competitors like AMD and Google's TPUs. These are Anthropic's own claims to investors, though, and have not been independently verified or benchmarked by outside parties.
- Anthropic says its Mythos Preview model cut AMD GPU onboarding time from up to a year to roughly three days.
- The model reportedly wrote almost all the kernel code needed, a task that traditionally required teams of engineers.
- Anthropic already treats Nvidia, AMD, Google TPU, and AWS Trainium chips as one interchangeable compute pool, including a new deal for up to 2GW of AMD capacity.
Source: tomshardware.com — Shane Downing, 2026-11-16
Published there as: “Anthropic says Claude Mythos Preview cut AMD MI355X onboarding from up to 12 months to around three days”
Read the original report → The summary and analysis above are GoKawiil's own, written from reporting by the source above. Facts and quotes belong to the original publisher.