AI News
Nov 13, 2026 · ThursdayPublic sources · Daily newsPublished daily at 08:00 Beijing time

AI News Daily

AI developments selected from public sources

Issue 20613Nov 2026Thursday
11monthThis month 16 issues
12
major stories
41
sources
5
first-party stories
About 7 min read

Lead story

Benchmark chart comparing Gemini 4 Hydrogen to other models

Google DeepMind Releases Gemini 4 Hydrogen for Everyday Coding and Agentic Workflows

Google DeepMind releases Gemini 4 Hydrogen, focused on speed and efficiency for everyday coding and agentic workflows; it scores 51 on the Artificial Analysis Intelligence Index at the default high thinking level, above Argon's 50, and can reach 54 at the xhigh level. The model's 1-million-token context window is priced uniformly at $1.50 per million input tokens and $7.50 per million output tokens, and it offers four thinking levels: low, medium, high, and xhigh. Google also launched Orchestrator in Antigravity, which can dynamically switch between Carbon and Hydrogen within the same session; on Terminal-Bench 4.0 it approaches Carbon's task success rate while consuming 24% less quota on average and running 19% faster.

First-party sourcesGoogle DeepMindAlso 15 other sources reported thisOriginal ↗
01

Model releases / updates

Box CEO tests Gemini 4 Hydrogen: accuracy 5 percentage points higher than Gemini 3.8 Flash

Aaron Levie said Box tested Gemini 4 Hydrogen during early access using Box AI Complex Work Eval; its accuracy was 5 percentage points higher than Gemini 3.8 Flash, typical tasks took less than half the time, and it used about 12% fewer tokens. The evaluation had Box Agent use Hydrogen to handle complex enterprise document tasks in industries including finance, consulting, and life sciences. Hydrogen will soon launch in Box AI Studio, where customers can use it to build custom agents.

Original

Baidu releases ERNIE 5.5 Flash Preview, 280B total parameters, 14B activated per token

Baidu released ERNIE 5.5 Flash Preview, with 280B total parameters and only 14B activated per token. Arena.ai said the model, with 1,668 points on Code Arena's WebDev leaderboard, is on track to become the second-ranked open-source model, and its weights will be open-sourced under Apache 2.0. Its score is slightly above DeepSeek-V4.1-Flash (Max)'s 1,664 and Kimi K3 (Max)'s 1,658, and comparable to Qwen3.8-Max's 1,670.

Original
02

Product release/Updated

Anthropic Launches Claude Meetings; Claude Can Join Video Meetings as a Participant

Anthropic released the beta of Claude Meetings, in which Claude joins video meetings as a visible participant, follows the conversation and simultaneously produces documents, whiteboards and meeting notes. Teams can ask questions by voice or chat during the meeting; Claude answers in text by default and, with permission, can call connected tools. The feature is first available on Google Meet for Team and Enterprise plans, does not join meetings with external participants by default, and notes and transcripts are free during the beta.

Original

Google Partners with Gentle Monster and Warby Parker to Launch Smart Glasses Equipped with…

Google announced that preorders have opened for new smart glasses developed with Gentle Monster, Warby Parker and Samsung, with an official release on December 10; the Gentle Monster series starts at $449 and is aimed at the US and South Korean markets. The glasses connect to Gemini and support hands-free operation, real-time voice translation (more than 70 languages), Google Maps navigation and Guided Vision audio descriptions for people with low vision. The hardware features an approximately 41-gram frame, a 12-megapixel ultra-wide camera, up to 9 hours of battery life, and a portable charging case that can recharge the glasses 7 times.

Original

Anthropic Launches Self-Hosted Claude Managed Agents Private Beta

Anthropic launched self-hosted Claude Managed Agents, currently in private beta, which can run agents in an enterprise's own cloud or on-premises environment. This version keeps the harness, sandbox, credentials and session history entirely in the customer's own environment; model requests can be sent to Anthropic, Amazon Bedrock, Google's Agent Platform, Microsoft Foundry or a self-built gateway, and session history is stored in the customer's own database. The company also described four security design features, including separation of the harness and sandbox, credentials not entering the sandbox, network access restricted by an allowed-host list, and full session records exported via OpenTelemetry.

Original
03

Industry news

Nvidia CEO Jensen Huang Resigns; CFO Colette Kress Becomes Interim CEO

The Nvidia board announced that Jensen Huang has resigned as president, CEO and board member effective immediately, and that Chief Financial Officer Colette Kress will serve as interim CEO while the board considers internal and external candidates. The resignation came four days after the company announced an internal investigation on November 8 into his relationship with Alice Huang; people familiar with the matter told WIRED that his relationship with a subordinate and business dealings with Megaspeed may have weighed more heavily on the board's decision than the Chinese espionage allegations. Jensen Huang co-founded Nvidia in 1993, and in summer 2025 the company became the first ever to surpass a $4 trillion market value; as of September 2026 it held about 90% of the GPU market.

Original

Anthropic Plans to Begin Public Trading on November 19; Retail Investors Can Subscribe for IPO Shares in Advance

Anthropic plans to begin public trading on November 19, with retail investors able to apply for IPO shares through participating brokers before pricing. Deutsche Bank gave a reference price of $70 to $120 per share, and X already supports the $ANT ticker.

Original

NVIDIA Board Discusses Whether to Remove CEO Jensen Huang; Shares Fall Over 8% After Hours

Reuters, citing people familiar with the matter, reported that the NVIDIA board met on the evening of November 11, 2026 to discuss whether to remove founder and CEO Jensen Huang; after The Wall Street Journal first reported that the company was preparing to fire him, the stock fell more than 8% after hours. On November 8, the board had stated its confidence in Jensen Huang and his management team and announced an independent investigation into reports that he had been targeted by a foreign intelligence agency. If the after-hours decline holds, NVIDIA's market value loss since the investigation statement would be about $600 billion.

Original

Nvidia Board Discusses Removing Jensen Huang, Musk Says He Is Willing to Take Him In

Some media reported that the Nvidia board was preparing to remove Jensen Huang, with a decision to be made on November 12, after another extramarital affair of his was exposed, this time with Kimberly Powell, a vice president of the medical business who reported to him. After the news broke, Nvidia shares fell more than 8% after hours, and a Blind survey of 232 employees on November 9-10 showed 88% supported Jensen Huang continuing as CEO. The article also reviewed potential successors including Cook, Su Zifeng, Rene Haas, Colette Kress and Ian Buck, as well as Musk's response on X to "please hire Jensen Huang," saying "that would be my honor."

Original

Stripe Completes Acquisition of AI Model Aggregation Platform OpenRouter, Deal Reportedly About $7.5 Billion

Stripe announced on November 10 that it had completed its acquisition of OpenRouter, an AI model API aggregation platform. It had announced agreement to the acquisition on August 19; The New York Times reported the deal value at $7.5 billion, while Bloomberg put it at more than $7 billion. OpenRouter already provides access to more than 400 models from over 80 providers and serves more than 10 million developers and enterprises. In October 2026 it processed about 790 trillion tokens, up 37% month over month, of which about 76% came from agent-related API calls. The two sides said they will work together to optimize token routing and usage to help enterprises use AI resources more effectively.

Original
04

Tips and perspectives

Character.AI files for bankruptcy; why are some AI companion products still profitable?

Character.AI filed for Chapter 11 bankruptcy protection in the United States on October 12 and plans to sell the company's assets under court supervision; if no buyer is found, it may shut down. The article compares its subscription-plus-free-unlimited-chat model with Zeta's and Tipsy's pivot to interactive storylines and virtual-currency payments, noting that the latter are already profitable; Zeta, for example, posted revenue of 70.9 billion won and operating profit of 5.2 billion won in the first half of this year.

Original
05

Briefs

End of issue

AI News DailyAutomaticallycompiledfrom public reports, with source links · View archive