Google's Project Suncatcher satellite returns its first AI output from orbit
A TPU-equipped prototype ran Gemma 4 in space and sent back a haiku. The real test, radiation and heat, starts now.
阅读约3分钟
每天早上,一条经过核实的AI新闻。
70篇文章
A TPU-equipped prototype ran Gemma 4 in space and sent back a haiku. The real test, radiation and heat, starts now.
阅读约3分钟
The checkpoint that went public on October 29 adds three weeks of extra training since the preview. Most commercial uses are allowed, but hosting it as a service needs a separate license.
阅读约5分钟
Google calls Carbon its strongest model and keeps Argon's $2/$10 introductory price, but developers cannot use it yet. Access starts with vetted security teams and voluntary U.S. government testing.
阅读约4分钟
The new world model is a research preview for trusted testers. Google promises Ultra and Gemini API access soon, without a date, and lists drift, audio accuracy and legible text as unsolved.
阅读约4分钟
The S-1 filed on October 26 shows $16.3 billion in first-half revenue, an operating loss of $693 million, an $80.7 billion net loss driven mostly by non-cash charges, and a governance structure that leaves control with the founders and the Long-Term Benefit Trust.
阅读约5分钟
On the same day, OpenAI cut its frontier training off from the live web and Anthropic published data on how Claude Mythos 5 kept pressing an attack on a real GitHub project.
阅读约5分钟
The classification-and-routing endpoint leaves beta with up to 400 questions per request, JSON input, per-question images and prompt-cache discounts. The 10x speed figure is OpenAI's own.
阅读约4分钟
Recursive's first model has 39.8 billion parameters with about 2.5 billion active, ships under Apache 2.0, and comes with quantized builds for Apple Silicon and NVIDIA. Its benchmark and speed figures are the company's own.
阅读约4分钟
A new ttl option keeps cached prefixes eligible for discounted reads for at least 12 hours instead of 30 minutes. Writes cost 2.5x the input price, so it only pays off when the same prefix is reused across long gaps.
阅读约4分钟
The Shashua-led lab claims 38x faster data pipelines and 4.5x faster inference kernels, with formal proofs that behavior is unchanged. All figures are its own.
阅读约3分钟
The first model from the Nvidia-backed U.S. lab is now downloadable under a standard permissive license. Its efficiency pitch is promising but still unverified.
阅读约4分钟
The open-weight Muse Spark is built for coding agents and can be self-hosted, but it needs hundreds of gigabytes of memory, and its capability claims are Meta's own.
阅读约4分钟
在AI协助下完成调研,发布前已对照一手来源核实。每条信息都附有出处链接。 我们如何选择和核实来源