Release4 sources

Google’s Gemini 4 coding agent now switches models mid-task — and finishes 19% faster

The New Stack · By Amanda Caswell · 2h ago

The Brief · written from 4 sources

Google released Gemini 4 Hydrogen, available globally in Search for all users with improved intent understanding and broader link discovery
The model handles multi-part questions, typos and highly specific requests better, breaking questions down and addressing each part
Artificial Analysis places Gemini 4 Hydrogen's Intelligence Index just below Claude Sonnet 5.5 and GPT-6.1 Sol, at about a third of Sonnet's cost but roughly 2.5x that of Sol

Why it matters

Developers building on Google Search, AI Studio or the Gemini API get a new default model with claimed better handling of complex queries, and the coding agent orchestration approach may change how multi-model coding tools are built.

Who’s affected

Developers using Google Search or AI Mode
Developers using AI Studio or the Gemini API
Teams evaluating coding agents

Gemini 4 Hydrogen
Gemini 4 Carbon
Claude Sonnet 5.5
GPT-6.1 Sol
#gemini
#google
#llm
#search
#coding-agents

Google launched Gemini 4 Hydrogen on Thursday alongside a new system that changes how coding agents use AI models. Rather than keeping its most capable model running across an entire session, Google is pairing Hydrogen with the more powerful Gemini 4 Carbon and switching between…

Read the full story at thenewstack.io ↗