International Edition
Latest News
Technology

Google Launches Gemini 4 Hydrogen to Replace Flash Models

Google released Gemini 4 Hydrogen on Tuesday, launching the new AI model with relaxed access limits across its ecosystem. The release replaces the Gemini 3.8 Flash model in company products, according to Ars Technica. Hydrogen is built to…

Google Launches Gemini 4 Hydrogen to Replace Flash Models

Google released Gemini 4 Hydrogen on Tuesday, launching the new AI model with relaxed access limits across its ecosystem.

The release replaces the Gemini 3.8 Flash model in company products, according to Ars Technica. Hydrogen is built to replace the faster, more economical Flash models that most users rely on.

Unlike previous releases such as Argon and Carbon, which rolled out slowly and restricted access to high-priced AI subscribers, Hydrogen access began immediately for a wider audience. Google claims the model runs significantly faster and reduces fabricated claims compared to older versions.

Hydrogen operates in four thinking levels, which include low, medium, high, and a new extra-high mode. Even when set to high mode, the model runs 2.5 times faster than Gemini 3.8 Flash, Ars Technica reported.

Google analysis indicates that Hydrogen reduces fabricated claims by more than five times.

The model also more than triples the score of 3.8 Flash on Terminal-Bench for code generation tasks.

In some benchmarks, Hydrogen matches or exceeds Argon. It scored 54 on the Artificial Analysis Intelligence Index using extra-high mode, while Argon scored 50.

Hydrogen stands on its own for everyday work or pairs with Carbon and Argon in multi-model software development workflows, blog.google reported. Development teams rely on frontier models like Carbon and Argon for deep architectural judgment while using Hydrogen as a fast execution engine for writing code and running tests.

Google introduced an Orchestrator tool in Google Antigravity to automate multi-model collaboration.

“[O]rchestrator dynamically switches between models like Carbon and Hydrogen within the same session,” according to Google.

When tested on Terminal-Bench 4.0, Orchestrator achieved success rates close to Carbon while consuming 24% less quota and finishing 19% faster. Carbon coordinates parallel agent workflows while Hydrogen powers worker and verifier agents across isolated workspaces, allowing teams to build and test changes simultaneously.

About the author: Anika Shah - Technology

MSc in Computer Science, senior reporter. Anika focuses on AI ethics, cybersecurity, and emerging hardware—frequently moderating panels at CES and Web Summit. “Anika Shah decodes tech breakthroughs and startup disruption shaping tomorrow’s digital landscape.”