IndustryMulti-source

OpenAI 曾洽购可解释性初创公司 Goodfire,报价超 50 亿美元

Why we picked it

OpenAI 想花 50 多亿美元买一家专门“透视模型内部”的公司,对方还是 Anthropic 投资的。如果你的模型推理藏在隐藏循环里,可解释性就是刚需。做安全、评估或关注思考链透明度的人都该看看。

The Information 报道,OpenAI 曾探讨以超过 50 亿美元的价格收购 Anthropic 支持的可解释性初创公司 Goodfire。这笔潜在交易的核心,是买下“看懂模型内部行为”的能力。

Body

Goodfire 的技术路线跳过模型的 chain of thought 输出,直接读取模型内部运行的数学表征,来判断模型为什么做出某个行为。这与 GPT-6 Astra 被曝在隐藏循环中完成部分推理、可读的思考链变少的背景直接相关——9 月曾有相关报道引发关于推理过程不可监控的争议。

此举也可视为 OpenAI 与 Anthropic 在可解释性方向上的竞争延伸:Anthropic 长期投入可解释性研究,Goodfire 获得 Anthropic 支持,OpenAI 若收购不成,等于竞争对手阵营多了一个对自家模型有审视能力的工具。

Original · kimmonismus

Interesting: OpenAI explored buying Goodfire, the Anthropic-backed startup that looks inside AI models to understand why they behave the way they do, for more than $5 billion, The Information reports.

You probably remember the uproar in September, when reports said GPT-6 Astra does part of its reasoning in hidden loops to be more efficient, leaving less readable chain of thought to monitor.

Goodfire's approach skips the chain of thought and looks at the math running inside the model directly.

Industry Coordinates & Source Verification

This report is categorized under AITOP's "Industry" desk, focusing on US frontier AI labs and Silicon Valley technology leaders. Within our cross-regional analytical framework, it directly maps to US-China foundation model benchmarks and commercialization comparison and AI developer toolchains, coding agents, and engineering workflows. Key entities and technical topics under continuous tracking include OpenAI, Goodfire, Anthropic, interpretability.

Source attribution: curated from kimmonismus (@kimmonismus). Cross-source corroboration: 11 independent sources have covered this development (including The Information, PolymarketMoney, IT之家), confirming industry-wide signal strength.