Anthropic 路演披露：Mythos 三天完成 Claude 在 AMD GPU 上的移植
Anthropic 自己爆料，让 Mythos 干 AMD GPU 适配，3 天搞定，Claude 自己写了近 100% 的内核，以前这活要几十个工程师干几个月。做底层工程和算力采购的人该看看，芯片生态的门槛可能要变了。
Anthropic 在路演视频中披露，今年首次接入 AMD GPU 平台时，让模型 Mythos 尝试移植工作。以往将 Claude 带上一个新的芯片平台需要数十名工程师投入数月，而 Mythos 在几乎没有人工干预的情况下，3 天就让 Claude 在 AMD GPU 上跑了起来。据称近 100% 的内核移植和编写工作由 Claude 自行完成。
这一案例展示了 Mythos 在底层系统工程任务中的实际能力：面对全新的硬件平台，它能自主完成内核级代码的编写与调试，工作量相当于过去一个工程师团队的数月产出。对 Anthropic 来说，这也降低了训练基础设施对单一 GPU 供应商的依赖，为在 NVIDIA 之外扩展算力供给提供了路径。
对行业而言，这暗示前沿模型已能承担芯片适配这类高门槛工程，AMD 等非 NVIDIA 硬件生态的接入成本可能大幅下降，依赖 CUDA 护城河的竞争格局值得关注。
A 社路演爆料，他们已经开始用 AMD 的 GPU Mythos 用了三天，就写出了 AMD 上跑 Claude 的内核。 以前这需要数十名工程师花费数月的时间。 Andrew Curran @AndrewCurran_ From the Anthropic Road Show video: 'This has also made us really good at bringing up new chips. As an example, this year, we started working with AMD GPUs for the first time. In the past, bringing on a new chip platform took dozens of engineers months of work. This time we had Mythos try it out. It got Claude running on AMD in just 3 days with basically no human intervention. Claude ported or wrote almost 100% of the kernels.' 🔗 View Quoted Tweet📊 6 ⚡
This report is categorized under AITOP's "Models" desk, focusing on US frontier AI labs and Silicon Valley technology leaders. Within our cross-regional analytical framework, it directly maps to AI developer toolchains, coding agents, and engineering workflows and US-China foundation model benchmarks and commercialization comparison. Key entities and technical topics under continuous tracking include Anthropic, Mythos, AMD, Claude.
Source attribution: curated from orange.ai (@oran_ge). Cross-source corroboration: 11 independent sources have covered this development (including Andrew Curran, Decoder, Simon Willison’s Weblog), confirming industry-wide signal strength.
