Google launches Gemini 4 Hydrogen to replace Flash, edging out Argon
Google says it scores one point above Argon on the Artificial Analysis index, at $1.50 per million input tokens and $7.50 per million output.
Models, from Google DeepMind
Google dominated the day, with a new Flash replacement and an Antigravity mode that assigns planning and edits to different Gemini 4 models. The rest of the news qualified the picture: an agent monitor that flags attacks only after the fact, reasoning that is getting harder to read, and a speech model beaten on speed and price.
7 stories, about 9 minutes to read them all.
Google says it scores one point above Argon on the Artificial Analysis index, at $1.50 per million input tokens and $7.50 per million output.
Models, from Google DeepMind
The 12 November release makes Code Mode an experimental opt-in and stops plugins outside their stated Claude Code version range from loading.
Coding tools, from Claude Code releases
Artificial Analysis ranks Scribe v3 near the top on accuracy, but Microsoft and xAI models run many times faster at about half the price.
Models, from The Decoder
The safety team's analysis of models from GPT-5 to GPT-6 Astra finds that capability, architecture, reinforcement learning and teacher models all push reasoning out of view.
Research, from The Decoder
Agent Anomaly Detection, in Private Preview, caught 4 of 18 poisoned-memory cases and flags after the fact; it does not block actions.
Safety and security, from Google for Developers
A Hugging Face post says the markup is now in both models, but the captured page gives no syntax, license or usage details.
Open source, from Hugging Face
The model option hands planning to Gemini 4 Carbon and edits to Hydrogen; Google reports 24% less quota use on Terminal-Bench 4.1.
Coding tools, from antigravity.google