Google released Gemini 4 Hydrogen on November 12, a model built for fast everyday coding and agent work. Google says it scores 51 on the Artificial Analysis Intelligence Index, one point above Gemini 4 Argon, at a lower price than Argon.
Google announced the model in a blog post from Koray Kavukcuoglu, SVP of Google DeepMind, and the rollout is global across developer, enterprise and consumer products. Google positions Hydrogen as the fast execution model, with Argon and Gemini 4 Carbon covering deep, long-horizon reasoning.
Gemini 4 Hydrogen Costs Less Than Argon
Hydrogen costs $1.50 per million input tokens and $7.50 per million output tokens, and the price stays the same across the full 1 million token context window. Argon launched at an introductory $2 and $10 that doubles later, and Gemini 3.7 Flash launched in August at an introductory $0.75 and $3.75, so Hydrogen sits between the two.
Cached input costs $0.075 per million tokens through implicit caching, which applies automatically to prompts of at least 4,096 tokens. Google AI Studio has a free tier with rate limits, and grounding with Google Search is billed per query. Hydrogen does not support explicit context caching.
Hydrogen offers low, medium, high and a new xhigh thinking level. High is the default setting, and at xhigh the Artificial Analysis score rises from 51 to 54. The xhigh level adds latency and token use and can overthink simple prompts, and the API rejects the minimal level.
Output Over 64K Needs Background Mode
Hydrogen can produce up to 1 million output tokens, but a standard request caps output at 65,536 tokens. Longer output needs background execution with a higher max_output_tokens value, and without that setting the output still stops at 65,536 tokens. Input, thinking and output share one 1,048,576 token context window.
The model takes text, images, audio and video files as input, and its API model ID is gemini-4-hydrogen. Google adapted it from Gemini 4 Carbon and gave it a knowledge cutoff of March 2026, per the model card.
Developers need to change a few API settings. Hydrogen ignores temperature, top_p and top_k if you send them, and tool calls have to be declared natively because raw-text JSON tool calls with no declared tools can return a MALFORMED_FUNCTION_CALL error.
Google Says It Runs 2.5x Faster
Google says Hydrogen finishes tasks about 2.5 times faster than Gemini 3.8 Flash at its default setting, and it cuts fabricated claims by more than 5x. Unnamed partners reported runs that were 2x to 2.8x faster with 12% to 23% fewer tokens on graph, SQL and data-heavy workflows. Every benchmark figure below comes from Google.
- 57% on Terminal-Bench 4.1, against 18.2% for Gemini 3.8 Flash
- 58.8% on HLE-Diamond, which leads its class
- 52.3% on the Vals Tax Agent Bench, ranked first overall
- 55.8% on the Vals Legal Research Bench, tied for first
- Fifth place on Agent Arena, and 71.6% on Agents’ Last Exam v1.1
Google Antigravity gets a new Orchestrator that switches between models such as Carbon and Hydrogen inside one session. On Terminal-Bench 4.0, Orchestrator reached near-Carbon task success while using about 24% less quota and finishing about 19% faster on average. Carbon can also coordinate parallel agents while Hydrogen runs the worker and verifier agents that build and test changes in isolated workspaces.
On safety, Google says Hydrogen cuts overreach in agentic tasks by up to 81% compared with Gemini 3.8 Flash, and it is up to 2x less likely to bypass permission boundaries when a task is blocked. It is also more than 2.3x more resilient to indirect prompt injection on the Gray Swan benchmark.
Google names hallucinations and occasional slowness or timeouts as known limitations. Token use also rises at higher thinking levels, and work on jailbreak resistance is still ongoing. Hydrogen did not reach the critical capability levels for chemical, biological, radiological, nuclear or cyber risks, but it did reach the alert thresholds for both, and Google says the full Safety and Security Report is coming soon.
Free Users Get It by Auto Selection
Hydrogen is rolling out to developers, businesses and consumers.
- Developers get it in the Gemini API and Google AI Studio on all tiers, in Android Studio on the paid tier, and in Google Antigravity with a paid Google AI Pro or Ultra subscription.
- Businesses get it in Gemini for business through the API or the model dropdown, in Antigravity, and in Gemini in Google Sheets with an eligible subscription.
- Consumers get it in the Gemini app model dropdown and in Google Search, and Google AI Pro and Ultra subscribers also get it in Gemini in Google Sheets.
In the Gemini app, Hydrogen follows instructions more closely and calls tools more reliably. It also pulls context from Gmail, Google Drive and connected apps such as Airtable and Dropbox. In Google Search, it handles typo-filled and multi-part queries better, and an upgraded query fan-out technique shows a wider range of links.
Free Gemini app users get Hydrogen through auto model selection when it fits the prompt. Google had already cut free users to Flash-Lite on October 9. In Sheets canvas, Google says outputs run over 30% faster than with previous models.
