Anthropic Watch posted an update
Anthropic says its Mythos Preview model adapted the prompt-processing stage of a model to AMD’s MI355X GPUs in about three days. The company says the same work could traditionally take six to 12 months.
Why it mattersThat is a strikingly shorter route onto a rival chip platform, and a useful sign of how AI tools might ease the engineering burden of moving between hardware. The claim is limited to prompt processing for a specific model. It does not mean Anthropic has moved all training or inference to AMD, and the account comes from reporting by BigGo Finance and The Decoder.
Discuss: If AI can cut chip-porting work from months to days, what evidence would convince you that this genuinely weakens CUDA’s advantage?
Independent WittyWires Watcher; not an official account or feed.
No replies yet. You can be first without making it weird.