Skip to content
World clockEU--:--UK--:--USA--:--CN--:--PLDEFRIT中文EN

portal about AI and technologyevents · analysis · interviews · technical background

Search
LIVE
›

AI & models

640 texts · 6/36
Meta seeks $20bn-$25bn bond sale to fund AI spending

Meta seeks $20bn-$25bn bond sale to fund AI spending

Meta is seeking to raise $20bn to $25bn in a bond sale on Tuesday, two-thirds in US dollars and one-third in euros, to fund AI spending, people familiar with the matter told CNBC; pricing was expected that afternoon.

AI & models4 November 20263 min readSources 28
OpenAI plans to buy Google TPUs, The Information reports

OpenAI plans to buy Google TPUs, The Information reports

OpenAI plans to buy Google's Tensor Processing Units, The Information reported, citing a person familiar with the matter, a step that would reduce its reliance on Nvidia. Neither company has confirmed the plan or disclosed terms.

AI & models4 November 20263 min readSources 28

Latest

12 texts
CERT-In uses Anthropic's unreleased Mythos to find software flaws

CERT-In uses Anthropic's unreleased Mythos to find software flaws

India's CERT-In has access to Anthropic's unreleased Mythos model and is using it to find vulnerabilities and recommend patches, government sources told CNBC-TV18, but cloud access raises concerns about sensitive data.

AI & models4 November 20263 min read
Statistics Canada study finds no broad evidence AI drove Canada's youth hiring decline

Statistics Canada study finds no broad evidence AI drove Canada's youth hiring decline

Statistics Canada found no broad evidence that AI led to substantial decreases in hiring or increases in job separation rates in Canada since ChatGPT launched in November 2022, according to a study published last week.

AI & models4 November 20264 min read
Sarvam AI releases Prism, an open-weight decision model with a 44 ms median response

Sarvam AI releases Prism, an open-weight decision model with a 44 ms median response

Sarvam AI released Prism on 3 November 2026. The open-weight model scores a fixed set of options instead of writing text, has a median response time of 44 milliseconds on one NVIDIA B200 GPU and scored 71.8 across three indices.

AI & models4 November 20263 min read
04

No Korean brokerage among 22 institutions named in Anthropic IPO filing

No Korean brokerage is among the 22 financial institutions named in Anthropic's US IPO filing, The Korea Herald reported, potentially leaving Korean retail investors without direct access to shares before the Nasdaq debut.

AI & models4 November 20263 min read
05

MIT researchers show GPT-6 Astra can rewrite other models' weights

MIT researchers have shown in a research blog that GPT-6 Astra can read and modify other models' weights without retraining. Changing 0.016 percent of Qwen3.5-0.8B's weights lifted its MATH-500 accuracy from 7.9 to 48.1 percent.

AI & models4 November 20264 min read
06

Tencent publishes Hy-Nomas technical report and 280-item benchmark

Tencent Hunyuan published the technical report and 280-item Hy-Nomas-Bench set for its Hy-Nomas world model on 30 October, giving researchers a fixed target for long-horizon video consistency. The arXiv paper appeared on 27 October; weights are downloadable under Apache 2.0.

AI & models4 November 20263 min read
07

Mistral AI releases weights for trillion-parameter Mistral Large 4

Mistral AI published the weights of Mistral Large 4, its trillion-parameter model, on 29 October 2026, letting companies and governments run it on their own hardware. The open release came 23 days after the model's public preview on 6 October.

AI & models4 November 20263 min read
08

MIT researchers say GPT-6 Astra can edit other models' weights

In a research blog, MIT researchers say OpenAI's GPT-6 Astra can read and edit other models' weights, a method called Weight-Space Programming that could enable targeted fixes without full retraining.

AI & models4 November 20263 min read
09

Docker urges isolation and governance to secure AI agents

In November 2026, Docker published guidance naming five security risks from AI agents that browse, run code and call tools with limited oversight, arguing isolation and governance, rather than model improvements alone, are the main defences.

AI & models4 November 20263 min read
10

Fal raises $1.3 billion Series F at $16 billion valuation

Fal has raised $1.3 billion in a Series F led by Bessemer Venture Partners, with Nvidia, Sound Ventures, HOF Capital and T. Rowe Price participating, valuing the AI inference startup at $16 billion, double its April valuation.

AI & models4 November 20263 min read
11

LiberAI unveils Liber-0, a world action model scoring 36.0% on 60 GPT-6 tasks

Beijing-based LiberAI has unveiled Liber-0, a world action model pretrained on more than 2.2 million hours of human experience, which scored 36.0% overall across 600 trials on 60 GPT-6-proposed tasks without robot demonstrations collected specifically for them.

AI & models4 November 20265 min read
12

Tencent releases Hy4 with cybersecurity training limited to defensive work

Tencent released its flagship Hy4 model on Wednesday 4 November 2026, trained with cybersecurity data limited to defensive work, and said it will publish the weights and a system card in the coming days.

AI & models4 November 20264 min read
page 6 / 36← previous1…45678…36next →