Google DeepMind launches Gemini 4 Hydrogen for faster coding

Google DeepMind launches Gemini 4 Hydrogen model

Google DeepMind announced Gemini 4 Hydrogen on 12 November 2026 as a new model aimed at everyday coding and agentic workflows. The model is offered through Google AI Studio, the Gemini API, Gemini for Business, the Gemini app, Google Search, and paid Google AI Pro and Ultra subscriptions in Google Antigravity¹.

Hydrogen scores 51 on the Artificial Analysis Intelligence Index at its default high‑thinking level and can reach 54 at the highest xhigh setting, surpassing Argon’s 50 score. It uses a flat pricing model of $1.50 per million input tokens and $7.50 per million output tokens across a 1 million‑token context window. In multi‑model workloads, the new Orchestrator feature can switch between Carbon and Hydrogen, achieving near‑Carbon success rates while using 24 % less quota and completing tasks 19 % faster on Terminal‑Bench 4.0. Hydrogen also finishes tasks 2.5 times faster than Gemini 3.8 Flash for standalone use¹.

Gemini 4 Hydrogen is designed for rapid execution in software development, complementing slower, more reasoning‑heavy models like Carbon and Argon. It enables developers to write code, run tests, and verify changes quickly, either alone or as part of a coordinated multi‑model workflow where Carbon coordinates and Hydrogen powers worker agents. This approach aims to reduce token costs while maintaining high task success, addressing the need for efficient, scalable AI assistance in everyday coding tasks¹.

“I am writing to share an important update.”

Anthropic goes public with strong growth metrics

The filing also reports a 15% Adjusted Operating Margin and a net loss of $23.3 billion, largely due to a non‑cash charge².

Claude Enterprise reached 50,000+ employee users, with over 85% of key‑function staff actively engaging the tool. Legal teams using Claude Code grew from about 50 seats to 700, a 14‑times increase, while developers saw a 151% rise in effective output per year².

Anthropic positions itself as a public benefit corporation, emphasizing responsible AI development and risk mitigation. Its financial disclosures highlight both rapid scaling of AI services and significant non‑cash accounting adjustments that inflate the reported loss².

Nvidia appoints interim CEO after Jensen Huang resignation

Nvidia announced on 12 November 2026 that co‑founder Jensen Huang resigned as chief executive and Board Member following an internal investigation. The board appointed Chief Financial Officer Colette Kress as interim CEO while it begins a search for a permanent replacement. Harvey Jones became the company’s first Chair of the board and communicated the changes to employees via email³.

The resignation followed allegations of multiple extramarital affairs involving Huang and a direct report, as well as claims of contact with an alleged Chinese spy, prompting the board’s investigation. Kress has served as Nvidia’s CFO for over 13 years and is described as a key architect of the firm’s growth and capital strategy, giving the board confidence in her ability to maintain commitments to shareholders, customers and partners. In his Statement, Jones emphasized Nvidia’s technical strength, employee base and partner ecosystem as the foundation for continued growth under Kress’s interim leadership³.

Muse can place calls for paid users

Meta’s consumer AI agent Muse now lets paid subscribers place phone calls to U.S. businesses on their behalf, a feature first tested in September and rolled out in November 2026^4,5. The capability is limited to users who pay for a Muse subscription, with plans starting at $20 per month for 500 million tokens per week or $100 per month for 3 billion tokens⁴. Muse’s chief AI Officer Alexandr Wang said the tool is intended to handle tasks people prefer not to make themselves, such as calling a cable company or a restaurant without online reservations^4,5.

The subscription‑based model provides token allocations that power Muse’s interactions, including the new calling ability, and heavier users can purchase larger token bundles for more frequent use⁴. Early adopters reported that Muse could schedule appointments, make reservations, and wait on hold, exemplified by a recreated call with a Volvo dealership shared by Meta employee Matt Schlicht⁵.

Muse, which surpassed 10 million users last month, is part of Meta’s effort to turn consumer AI into a mainstream personal‑assistant product. As phone‑based customer service remains cumbersome, offering an AI that can navigate call trees and wait on hold provides a convenience that mirrors traditional human assistants, albeit with the risk of businesses rejecting AI callers⁴.

TSMC invests $1 billion in Anthropic and OpenAI

Taiwan chipmaker TSMC approved a $1 billion investment in AI developers Anthropic and OpenAI, allocating up to $500 million to OpenAI preferred stock and $500 million to Anthropic common shares at its expected IPO price. The board decision was reported on 11 November 2026. This move gives TSMC direct equity stakes in two leading AI firms⁶.

The investment splits evenly between the two companies, with each portion earmarked for preferred or common shares as described in the board plan. Analysts note that such circular financing makes it harder to gauge true market demand for AI hardware, even as TSMC reports six straight quarters of record profit, including more than $23 billion in the July‑September 2026 quarter⁶.

Circular financing, where chipmakers, AI developers and investors trade equity and hardware, has become common during the AI boom. Nvidia, for example, has already placed $30 billion in OpenAI and $10 billion in Anthropic, highlighting the scale of such arrangements. TSMC’s cash‑rich position, driven by sustained AI‑chip demand, enables it to join this trend despite recent volatility in tech stocks⁶.