Signal

Google launches Gemini 4 Hydrogen, calling it a major upgrade over Gemini 3.8 Flash, at $1.50/1M input and $7.50/1M output tokens across its 1M context window

First reported by Blog.google ·

The signal ●●●● Compiled by AI from Blog.google, Techmeme, Hacker News, TechCrunch, Ars Technica and 7 more
Why you might care

You can now use AI models that perform complex coding and agentic tasks significantly faster and more cheaply.

What happened

Google DeepMind has launched Gemini 4 Hydrogen, a new AI model designed for speed and efficiency in agentic workflows. This model is positioned as a significant upgrade over Gemini 3.8 Flash, offering faster task completion and improved intelligence. Gemini 4 Hydrogen boasts a 1 million-token context window and is priced at $1.50 per million input tokens and $7.50 per million output tokens. The model aims to enhance developer and enterprise workflows by performing iterative coding, debugging, and open-ended agentic tasks more effectively. It also features strong multimodal capabilities, synthesizing information from dense documents, spreadsheets, charts, audio, and video. Hydrogen's architecture supports adaptable speed and intelligence across different thinking levels, making it suitable for both standalone use and integration into multi-model workflows through Google's new Orchestrator feature in Google Antigravity.

What it means

Gemini 4 Hydrogen is engineered to be a more efficient execution engine, particularly in multi-model workflows where it can dynamically switch between models like Carbon and Hydrogen within a single session. This capability, facilitated by Google Antigravity's Orchestrator, allows for near-frontier model success rates at reduced quota consumption and faster completion times. The model's ability to handle iterative coding, debugging, and large repository analysis with focused edits and parallel processing aims to streamline software development cycles and reduce costs.

The introduction of Gemini 4 Hydrogen signifies a market trend towards specialized, cost-effective AI models optimized for specific use cases, moving beyond a one-size-fits-all approach. Its performance benchmarks across developer and enterprise workflows, including iterative coding, open-ended agentic execution, and multimodal knowledge synthesis, highlight its competitive positioning. The pricing structure, consistent across its large context window, suggests a move towards more predictable AI operational expenses for businesses integrating these advanced capabilities.

AI-written summary. May contain errors.

Google Crypto