模型发布/更新
-
OpenAI 发布 GPT-6 Astra:1.05M 上下文的计算机操作模型,因触及 Critical 网络安全阈值而限制访问
OpenAI 发布 GPT-6 Astra,定位为计算机操作模型,提供 1,050,000 token 上下文窗口、128,000 最大输出 token,2026 年 4 月 30 日知识截止,OSWorld V2-Offline 得分…
-
OpenAI releases new Astra model focused on computer and browser operation, sparking debate over opaque recurrence
OpenAI released its latest model Astra, calling it its most powerful model to date and highlighting computer and…
-
OpenAI 开始发布 GPT-6 Astra,面向全部 Plus 用户开放
OpenAI 宣布开始发布 GPT-6 Astra,称正以尽可能谨慎和快速的方式推进,重点让全部 Plus 用户可用,而不只限 Pro、Business 和 Enterprise 套餐。发布需几天完成,背后多个全新系统将首次大规模运行,团…
-
OpenAI 发布 GPT-6 Astra,多项基准达到 SOTA
OpenAI 发布 GPT-6 Astra,在 FrontierMath Tier 4、ARC-AGI 3、TerminalBench-4.0 上达到 SOTA,并在 Terminal-Bench Science 0.1 和 Health…
-
OpenAI 发布 GPT-6 Astra,首个达到关键级网络安全能力门槛的模型
OpenAI 于 9 月 3 日发布新一代大语言模型 GPT-6 Astra,是其首个达到准备框架中关键级网络安全能力门槛的模型,可在无逐步指导下发现防护严密系统的未知漏洞。
产品发布/更新
-
Hugging Face releases open-source tool funes, giving coding agents a locally held memory layer
Hugging Face released funes, an open-source tool that provides a local memory layer for coding agents such as Claude…
-
xAI designs Grok Bot to rebuild interface for persistent agents
xAI published a design article on how Grok Bot designs interfaces for persistent agents that go beyond single…
-
OpenAI launches Daybreak for Frontline Defenders with $1 billion to support frontline cyber defense
OpenAI announced the global Daybreak for Frontline Defenders program, pledging $1 billion in subsidized Daybreak…
-
xAI launches Grok Bot for enterprise, free for Grok and Cursor Enterprise customers for two weeks
xAI announced that Grok Bot is open to enterprises, with Grok and Cursor Enterprise customers able to use it free for…
行业动态
-
NVIDIA announces $12.9303 billion acquisition of Hugging Face
NVIDIA announced that it has agreed to acquire Hugging Face for $12,930,300,000, with Jensen Huang announcing the news…
技巧与观点
-
Artificial Analysis 评测 GPT-6 Astra:编码智能体追平 Fable 5 但价格涨至 2.5 倍
Artificial Analysis 发布 GPT-6 Astra 评测,其 Coding Agent Index 得分 67,约等于 Claude Opus 5 和 Fable 5,且成本不到 Fable 5 的一半;token 效率…
-
François Chollet reviews GPT-6 Astra performance on ARC-AGI-3
François Chollet wrote that GPT-6 Astra delivers a step-change in interactive reasoning, scoring 66% on ARC-AGI-3 with…
-
Rohan Paul breaks down safety findings in OpenAI GPT-6 Astra's 117-page system card
Rohan Paul summarized key points from OpenAI's 117-page system card for GPT-6 Astra: Astra's ability to control its…
-
Gary Marcus reviews GPT-6 Astra: clear progress but robustness and monitorability in doubt
Gary Marcus published a review of GPT-6 Astra, saying multiple reports show it is genuine progress and that OpenAI's…
-
Tom Tunguz analyzes data-to-compute logic behind Meta Muse Spark dual-tier pricing
Tom Tunguz analyzes Meta's release of the Muse Spark model and its dual-track API pricing: the Standard Tier…
-
Google Cloud shows how to run a persistent agent on Cloud Run instances for $5.70 per month
Shir Meir Lador explains on the Google AI developer blog how to run a persistent agent 24/7 in the cloud using Cloud…
-
Meta Muse Spark 1.3 在 Artificial Analysis 编码智能体指数中与 Claude 组合对比评测结果公布
Artificial Analysis 编码智能体指数显示,Meta Muse Spark 1.3 (max) 在 Muse Code 下得 68 分,仅次于 Claude Code + Opus 5 (xhigh) 的 68 分。
-
ARC-AGI-3 saturated by Astra just half a year after release, twice as fast as François Chollet expected
Sherwin Wu said he once found ARC-AGI-3 difficult, but the benchmark has now been saturated by Astra. Citing François…