ModelsMulti-sourceSelected

Claude Fable 5.2 生成可制造 PCB 设计,三款硬件原型直接投产

Why we picked it

硬件设计里最烦的 PCB 布局,这回模型自己会返工了。以前旧模型留一堆挡走线的元件没人修,Fable 5.2 自己重布几十次还能省三分之二成本。做硬件原型、小批量投产的团队可以关注 Diode 的实测数据。

Anthropic 的 Claude Fable 5.2 从简短提示词出发,为一部免提电话、一个语音助手中枢和一台掌上电子宠物产出了完整硬件方案,包括规格书、电路原理图、板级布局、可 3D 打印外壳和固件。板卡制造商 Diode Computers 直接按模型输出文件制造了这三款硬件,跳过了以往多轮人工修改制造文件的流程。

Body

技术上的关键差异在于布局返工能力。过去一年的旧模型在 Diode 手里只能做到 95% 完成度的布局,但往往因为不会撤销阻挡走线的元件摆放而无法投产。Fable 5.2 会在无人提示的情况下主动重做设计:把语音中枢的走线重布了几十次,并重新分配免提电话的芯片引脚,让触控走线互不交叉。

在 Diode 内部的 PCB Gym 布局基准上,Fable 5.2 得分 69%,Fable 5 为 25%,同时 API 成本约为前者的三分之一。硬件设计与制造的自动化程度由此明显提高,对硬件创业团队和小批量生产方是直接的效率变化。

Original · rohanpaul_ai

Claude Fable 5.2 took a speakerphone, a voice-assistant hub and a handheld virtual pet from short prompts to working prototypes.

From short prompts, it produced specs, schematics, board layouts, 3D-printable cases and firmware for a speakerphone, voice-assistant hub and handheld virtual pet.

Diode Computers, a board maker, built all 3 straight from the model’s files, skipping the usual rounds of manufacturing changes.

Over the past year, earlier models left Diode with layouts 95% complete yet unusable, because they rarely undid placements that blocked the wiring.

Fable 5.2 reworked its designs unprompted, rerouting the hub’s tracks dozens of times and reassigning chip pins so the speakerphone’s touch-control tracks never crossed.

On Diode’s internal PCB Gym layout benchmark, it scored 69% against 25% for Fable 5 at about 1/3 of the API cost.

Industry Coordinates & Source Verification

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 Claude Fable 5.2, Anthropic, PCB设计, Diode Computers.

Source attribution: curated from rohanpaul_ai (@rohanpaul_ai). Cross-source corroboration: 11 independent sources have covered this development (including kimmonismus, 宝玉, Artificial Analysis), confirming industry-wide signal strength.