Anthropic / Wikimedia Commons (Public domain)
The company says its AI did most of the work of porting Claude to AMD's MI355X chips, with minimal human help
Moving a large AI model onto new hardware usually takes months of engineering. Anthropic says its AI handled most of that job itself.
According to Anthropic, Claude Mythos enabled Claude to run on AMD’s MI355X GPUs in about three days, with minimal human help.
What Anthropic says happened #
The work traces back to testing in July 2026. An Anthropic engineer reportedly connected Claude to the new AMD hardware and told it to get going.
From there, the model handled the software migration and tuned performance across roughly three days. The research describes steadily improving performance curves, achieved without ongoing human intervention. AMD CEO Lisa Su reportedly confirmed that the assistance AMD offered turned out to be unnecessary.
The hardware and the model #
The MI355X is built on AMD’s CDNA 4 architecture. It carries 288 GB of HBM3E memory and 8 TB/s of bandwidth, with up to 10 PFLOPS of peak performance.
AI, tech, and the markets they move—in one daily briefing.
Daily. Free. Join 34,000+ readers across crypto, finance, and policy.
On November 9, 2026, Anthropic unveiled Claude Mythos 5.2, which it describes as its most advanced model yet in cybersecurity and life sciences.
Mythos is not something anyone can sign up for on a whim. Access runs exclusively through vetted programs, and Mythos-class models have shown significant capabilities in autonomous cybersecurity tasks.
Pricing for Claude Mythos is set at $8 per million input tokens and $40 per million output tokens. A fast mode doubles that cost.
The bigger AMD and Anthropic deal #
On July 22, 2026, AMD and Anthropic announced a strategic partnership built around AMD’s Instinct MI450-series GPUs.
Under that agreement, Anthropic commits to using up to 2 GW of MI450-series GPUs in AMD’s Helios rack-scale systems. Deployment is planned to begin in the first half of 2027.
AMD pledged an equity investment of up to $5 billion in Anthropic.
The partnership also includes a plan to use Claude to improve ROCm, AMD’s software stack for running AI workloads on its chips.
Disclosure: This article was edited by Diego Almada Lopez. For more information on how we create and review content, see our Editorial Policy.