DEV Community

News
News

Posted on

Google launches Gemini 4 Hydrogen, its new everyday speed-tier model, for everyone

Infographic cover

πŸ“… November 12, 2026 β€’ AI Automation

November 12, 2026. Google introduced Gemini 4 Hydrogen, a new model built for rapid execution across everyday coding and diverse agentic workflows. It joins the Gemini 4 family alongside the frontier models Carbon and Argon, which Google reserves for deep, long-horizon reasoning on the hardest problems. Hydrogen is the everyday workhorse of the lineup, and it is available from day one across all tiers of Google AI Studio, the Gemini API, Gemini for business, the Gemini app and Google Search, while paid Google AI Pro and Ultra subscribers get it inside Google Antigravity.
Free Gemini app users meet Hydrogen through automatic model selection whenever it fits the prompt. Pricing is flat across the entire 1 million-token context window: 1.5 dollars per million input tokens and 7.5 dollars per million output tokens.

On the Artificial Analysis Intelligence Index Hydrogen scores 51 at its default high thinking level and 54 at the highest xhigh setting, surpassing Argon at 50, while often completing tasks faster than models in its class. Four thinking levels β€” low, medium, high as the default and the new xhigh β€” let teams trade speed against reasoning depth on each task.
For teams migrating from Gemini 3.8 Flash the gains are concrete: tasks finish 2.5 times faster at the high default setting, Terminal-Bench 4.1 scores more than triple to 57 from 18.2 percent, fabricated claims are cut more than 5 times, and the model adapts cleanly when a tool call fails or a search returns empty. Hydrogen also posts 53.4 percent on the new Multimodal Knowledge Agent (MKA) benchmark for agentic knowledge work over complex visual files and 71.6 percent on Agents Last Exam v1.1 for end-to-end professional workflows.
Partner deployments ran 2x to 2.8x faster while reducing token usage by 12 to 23 percent compared to 3.8 Flash.

For multi-model work Google also introduced Orchestrator in Google Antigravity, which dynamically switches between models like Carbon and Hydrogen within the same session. On Terminal-Bench 4.0 it reached near-Carbon task success rates while using 24 percent less quota on average and completing tasks 19 percent faster.
The safety and security stack is equally detailed: overreach in agentic tasks is reduced by up to 81 percent and under-service by up to 79 percent, misaligned actions and task shortcutting are cut 12 times, and the model is up to 2 times less likely to bypass permission boundaries on impossible or blocked tasks.
Google reports a 96.7 percent honesty rate on its internal agent honesty evaluation, 99.9 percent chain-of-thought legibility with zero deceptive reasoning edits across more than 200000 audited turns, and over 2.3 times more resilience against indirect prompt injection on the Gray Swan benchmark. For automation builders the practical read is simple: a fast agentic tier for everyday work, plus automatic multi-model orchestration when a task needs the heavier models.

Sources: deepmind.google β€” 9to5google.com β€” jurelist.com

Key figures

IMPACT checklist


Continue the conversation: https://t.me/AIAutomationArchitecture Β· Bluesky

AI Automation

Top comments (0)