Models86°

OpenAI 发布 GPT-6.2,Anthropic 以每股 83 美元完成 IPO

Why we picked it

两家公司一周内动作密集:OpenAI 降价推 GPT-6.2 还给 Responses API 打折,Anthropic 上市当天就抛出黎曼猜想证明和黑洞物理结果。开发者该看 API 定价和 Code Mode,关注前沿研究的可以看那篇 Lean 形式化证明。

OpenAI 本周上线 GPT-6.2 Astra 和 GPT-6.2 Sol,覆盖 ChatGPT Work、Codex 和 API,其中 GPT-6.2 Astra 定价比 GPT-6 Astra 低 20%。Anthropic 则以每股 83 美元定价在纳斯达克上市,代码 ANT,同时完成 260 亿美元的私募配售。

Body

产品层面,ChatGPT 开始推送 Live Dictation、ChatGPT for Sales 与 Accounting,Space 功能向 Plus 用户全球开放;开发者侧 GPT-Live-1.1 进入 API,Responses API 的后台模式新增 2 小时和 8 小时的折扣完成窗口。Anthropic 方面 Claude Code 推出实验性 Code Mode,移动端应用也启动改版推送。

研究进展上,Anthropic 的 Levent Alpöge 宣称在与 Claude 协作下找到黎曼猜想的证明,目前仅以 Lean 形式化形式发布;Claude Fable 5.2 被报告解决了两个黑洞物理的开放问题。OpenAI 还与 Apollo Research、METR、Transluce、Goodfire 签署嵌入式评估协议,并披露其智能体可自主完成 46% 的 2-4 小时研究任务。

Original · Tibor Blaho

OpenAI and Anthropic this week: GPT-6.2, Anthropic IPO, Riemann Hypothesis claim, black holes, ChatGPT Space (Week 47, 2026)

OpenAI launched GPT-6.2 Astra and GPT-6.2 Sol in ChatGPT Work, Codex and the API, with Astra priced 20% lower than GPT-6 Astra

From my own testing, GPT-6.2 Astra upgraded a multimillion-line legacy app with all tests passing after about six hours

OpenAI started a science blog, opening with a candidate lattice definition of the Standard Model built with GPT-6 Astra

OpenAI signed agreements for a first phase of embedded assessments with Apollo Research, METR, Transluce and Goodfire, and shared that agents now complete 46% of two-to-four-hour research tasks at OpenAI on their own

ChatGPT is getting Live Dictation, ChatGPT for Sales and Accounting launched, and Space is rolling out globally to Plus users

For developers, GPT-Live-1.1 is available in the API, and the Responses API background mode got discounted two-hour and eight-hour completion windows in public beta

OpenAI is bringing Daybreak to NATO and allied cyber defense agencies, and signed journalism partnerships with LVMH's Media Division and Le Groupe Figaro

Anthropic priced its IPO at $83 per share and started trading on Nasdaq as ANT, then closed the offering with the full underwriter option and a $26 billion private placement

Anthropic's Levent Alpöge claims a proof of the Riemann Hypothesis found working with Claude, published for now only as a Lean formalization with a writeup still to come

Anthropic reported that Claude Fable 5.2 solved two open problems in black hole physics

Claude Code got an experimental Code Mode, linked billing launched in beta on the Claude Platform, and Anthropic started rolling out a refreshed Claude mobile app, and more

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 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 GPT-6.2, OpenAI, Anthropic, Claude.

Source attribution: curated from Tibor Blaho (@btibor91).