Gemini 4 Hydrogen is a fast, efficiency-focused model adapted from Gemini 4 Carbon, intended for everyday coding and agentic work rather than the deepest, longest-running reasoning tasks assigned to Gemini 4 Carbon and Argon. It accepts text, images, audio, and video, with a combined input-output context window of up to one million tokens; synchronous responses can reach 64,000 tokens, with longer outputs available through background decoding. Its target uses include repository coding and debugging, web development, multimodal knowledge work, research, writing, prototyping, and acting as a sub-agent alongside larger models. It is offered through the Gemini app and API, Google AI Studio, Sheets, business products, Search, and Antigravity. Known limitations include hallucinations, occasional delays or timeouts, potentially high token use at elevated thinking settings, and a March 2026 knowledge cutoff.
November 2026 benchmark results show strong but uneven performance against Claude Sonnet 5.5 and GPT-6.1 Terra. Hydrogen leads the listed comparisons on Agents’ Last Exam, AutomationBench-AA, Vending-Bench 2, several multimodal measures, and HLE-Diamond, but trails Sonnet on Terminal-Bench, OSWorld, and Terminal-Bench-Science. Its scores include 76.1% on AutomationBench-AA, $14,549 on Vending-Bench 2, and 58.8% on HLE-Diamond. Internal automated safety tests found better multimodal content safety than Gemini 3.8 Flash, while text and multilingual safety remained similarly high. Refusal tone was slightly worse, and unjustified refusals increased, especially for multimodal prompts. A full safety and security report was still forthcoming; the model card points to Carbon’s documentation for training, architecture, and policy details.
Summary generated by AI from the linked article. hn.today is not affiliated with Hacker News or Y Combinator.
A Hacker News poster asks which practical AI skills individuals can learn online to benefit economically from LLMs and agents. They want specific, hands-on recommendations suited to someone with a moderately technical background, without requiring a degree. (news.ycombinator.com)
Former OpenAI researcher Will Depue claims hidden initialization patterns in six Chinese open-weight AI models match Nvidia Blackwell hardware, including systems restricted from export to China. He says the matches were reproduced on real chips and independently checked by researchers, though the claims are based on his forensic analysis. (twitter.com)
Antithesis built a system that beats NetHack on nearly 80% of game seeds within five minutes, across character builds. Successful runs take a median of under two minutes, advancing the company’s goal of creating smarter fuzzers and autonomous systems. (antithesis.com)
OpenCyclone is a free, open-source Android agent that uses your existing apps to carry out tasks through a hidden screen. It asks for approval before actions like sending messages and lets users connect their own AI models. (opencyclone.vercel.app)
Sam Bankman-Fried’s $500 million investment gave Alameda nearly 14% of Anthropic, a stake that would be worth about $37 billion at the AI company’s expected $1.5 trillion IPO valuation. The stake was sold in bankruptcy to repay FTX customers and creditors after Bankman-Fried’s businesses collapsed. (ft.com)
llama.cpp and ggml have reached v1.0, establishing stable public C APIs and clear versioning rules for breaking changes. The release strengthens the foundation for running local, multimodal AI across a wide range of hardware. (huggingface.co)
Today's best Hacker News stories, summarized and screenshotted, one email a day.