Becky Simms Avatar

Written by Becky Simms

Contributor

SHARE
A glowing hydrogen atom symbol with orbiting electrons above a circuit board landscape, representing Google's new Gemini 4 Hydrogen AI model

Google has a new fast AI model, and it is already in front of everyone. On 12 November 2026, Google launched Gemini 4 Hydrogen, the third member of the Gemini 4 family, built for speed and everyday work rather than the deepest reasoning. It replaces Gemini 3.8 Flash as Google’s fast tier, and Google says it finishes tasks 2.5 times faster while being noticeably smarter. Unlike its bigger siblings, Hydrogen is rolling out to free users too, inside Google Search and the Gemini app.

The launch was announced by Koray Kavukcuoglu, senior vice president at Google DeepMind and chief AI architect. It comes only weeks after Google rolled out the two frontier models in the family, Gemini 4 Carbon and Gemini 4 Argon, and it completes the lineup: two heavy models for hard problems, one fast model for everything else. This article sets out the numbers, the pricing, the four thinking levels, where you can actually use Hydrogen today, and the honest limits in Google’s own benchmark claims.

The Numbers in One Place

MeasureGemini 4 HydrogenWhat it means
Speed vs Gemini 3.8 Flash2.5x fasterAt the default high thinking level, tasks finish two and a half times faster
Terminal-Bench 4.1 score57% vs 18.2% for 3.8 FlashMore than triple the agentic coding score of the model it replaces
Artificial Analysis Intelligence Index51 at high, 54 at xhighEdges out the larger Gemini 4 Argon, which scores 50
Pricing$1.50 per million input tokens, $7.50 per million output tokensFlat across the full 1 million-token context window
Context window1 million tokens in, up to 1 million outMatches the rest of the Gemini 4 family
Thinking levelsFour: low, medium, high, xhighHigh is the default; xhigh is new
Fabricated claims vs 3.8 FlashCut by more than 5xFewer invented facts in agentic work, according to Google
Partner testing vs 3.8 Flash2x to 2.8x faster, 12% to 23% fewer tokensEarly testing on complex graphs, SQL and data-heavy enterprise workflows

The figures come from Google’s announcement, cross-checked against WinBuzzer, MakeUseOf and The Hindu BusinessLine.

What Gemini 4 Hydrogen Is

Gemini 4 now comes in three tiers. Carbon and Argon are the frontier models, built for deep, long-horizon reasoning on the hardest problems. Hydrogen is the execution tier: the model for writing code, running tests, verifying changes, doing research and carrying multi-step tasks to completion. Google describes it as a standalone upgrade for teams still on Gemini 3.8 Flash, and as the high-efficiency engine inside multi-model workflows, where Carbon or Argon handle the architectural judgment and Hydrogen does the fast work.

That pairing is now partly automatic. Google is introducing Orchestrator in Google Antigravity, which switches between models like Carbon and Hydrogen inside the same session. On Terminal-Bench 4.0, Google reports Orchestrator reached near-Carbon task success rates while consuming 24% less quota and finishing 19% faster. For parallel work, Carbon can act as coordinator while Hydrogen powers worker and verifier agents across isolated workspaces.

Hydrogen is adapted from Carbon, according to its model card as reported by WindowsForum, which keeps the family consistent: all three models share a unified 1,048,576-token window and return text only.

Speed and Benchmark Results

The headline claim is speed. Google says Hydrogen finishes tasks 2.5 times faster than Gemini 3.8 Flash at its default high setting. The benchmark behind the launch is Terminal-Bench 4.1, a test of agentic terminal work, where Hydrogen scores 57% against Flash’s 18.2%, more than tripling the result. On the Artificial Analysis Intelligence Index, Hydrogen scores 51 at its default high level, edging out Argon at 50, and reaches 54 at its highest xhigh setting.

The wider benchmark picture, all reported by Google, is strong in the places that matter for everyday agentic work. Hydrogen ranks number 5 on Agent Arena’s live benchmark for multi-step tasks such as deep research, data analysis and codebase navigation. It leads its class on HLE-Diamond at 58.8%, ranks number 1 overall on the Vals Tax Agent Bench at 52.3%, ties for number 1 on the Vals Legal Research Bench at 55.8%, and leads its class on AutomationBench-AA. For knowledge work over complex visual files, it scores 53.4% on the Multimodal Knowledge Agent benchmark and 71.6% on Agents’ Last Exam v1.1.

The honest caveat is that these are Google’s numbers, measured on benchmarks Google chose. One independent reading of the model card shows the limit clearly: on Terminal-Bench 4.1, Hydrogen’s 57.0% sits below Claude Sonnet 5.5 at 63.9%, even though Hydrogen beats the same model comfortably on SWE-sweep, 13.5% against 3.1%, and against GPT-6.1 Terra at 5.8%. Hydrogen is the fastest model in its class, not the best coding model on every test. Independent verification will come as third-party evaluations publish their own results.

Pricing: What Hydrogen Costs

Hydrogen costs $1.50 per million input tokens and $7.50 per million output tokens, flat across its entire 1 million-token context window. There is no price increase for longer contexts, which is the part developers will notice most in real workloads.

ModelInput per million tokensOutput per million tokensNotes
Gemini 4 Hydrogen$1.50$7.50Flat across the full 1M window; free tier available
Gemini 3.8 Flash$0.75$3.75Through 31 December 2026, then rises to Hydrogen’s rates on 1 January 2027
Gemini 4 Carbon$2.00$10.00Introductory rate ending 30 November 2026, when both figures double to $4.00 and $20.00
Gemini 4 ArgonHigher than CarbonHigher than CarbonThe top frontier tier; no free tier

The comparison figures for Flash and Carbon come from WinBuzzer and OrcaRouter.

Hydrogen is the cheapest of the three Gemini 4 models, the only one with a free tier, and the only one Google ships to Search users worldwide. That pricing gap is the point: Google is not selling Hydrogen as the smartest model, it is selling it as the model cheap and fast enough to use for everything.

Four Thinking Levels, and Why They Matter

Hydrogen ships with four configurable thinking levels: low, medium, high and xhigh. High is the default, and xhigh is new to this model. The idea is simple: you only wait and pay for the reasoning depth a task actually requires. A quick rewrite or a simple question runs at low, a multi-step coding job runs at high or xhigh.

The levels scale in a smooth progression, according to Google, so the trade between quality, latency and cost stays predictable. The xhigh setting is where the benchmark peak sits: 54 on the Artificial Analysis Intelligence Index, against 51 at the high default. For most everyday use, the default high setting is the one that matters, and it is the setting behind the 2.5x speed claim.

Where You Can Use It Today

Hydrogen is rolling out globally from 12 November 2026 across Google’s developer, enterprise and consumer surfaces:

  • For developers: the Gemini API, Google AI Studio on free and paid tiers, and Android Studio.
  • For enterprises: Gemini for business through the API or the model dropdown, Google Antigravity, and Gemini in Google Sheets for eligible business subscriptions.
  • For everyone: the Gemini app and Google Search for all users, with Gemini in Sheets starting with Google AI Pro and Ultra subscribers.

In the Gemini app, paid users can pick Hydrogen from the model picker, where it is labelled “Everyday help”. Free users cannot pick it directly. Instead, they get it automatically when Google’s system decides it suits the prompt, which means free users are now running a Gemini 4 model for the first time. Free accounts lost access to Flash and Pro on 9 October, so Auto is now their route to Hydrogen.

Google says it is also rolling Hydrogen into Search for all users worldwide, which makes it the model behind the answers most people will see first. The company has not published a release date for each product, and says the rollout is gradual.

What Changes in Search and the Gemini App

For consumers, Hydrogen is meant to make everyday help faster and more reliable. Google says it stays on track and follows instructions more closely across multi-turn conversations, sticks closely to your tone, format and specific rules, and pulls together clear tables and visual breakdowns without dropping key details. It also calls the right tools and skills more often, drawing on uploads, Gmail, Google Drive and connected apps such as Airtable and Dropbox.

In Search, Google says Hydrogen brings three changes: it understands the true intent behind a question, even with multiple parts, typos or very specific requests; it gives more visual, easy-to-scan responses; and it upgrades query fan-out, connecting you to a broader range of relevant links than before. Google published side-by-side comparisons of Search results under Gemini 3.5 Flash and Gemini 4 Hydrogen for apartment repairs, Boston neighbourhoods and sneakers, showing the newer model giving fuller, more visual answers.

Safety: Built for Agentic Work

Hydrogen brings Google’s latest Gemini 4 safety and security stack to its broadest audience yet, and the safety claims are aimed squarely at agentic work, where a model acts on its own. Google reports Hydrogen reduces overreach in agentic tasks by up to 81% and under-service by up to 79%, and cuts misaligned actions and task shortcutting across tool-use and coding evaluations by 12 times. When faced with impossible or blocked tasks, it is up to 2 times less likely to bypass permission boundaries, and it records a 96.7% honesty rate on Google’s internal agent honesty evaluation.

The model also preserves 99.9% chain-of-thought legibility with zero deceptive reasoning edits across more than 200,000 audited turns, and Google says it is more than 2.3 times more resilient to indirect prompt injection attacks on the Gray Swan IPI benchmark. For everyday use, the safeguards are tuned to distinguish real biosecurity risks from legitimate medical or scientific questions, keeping false refusals on HealthBench Pro under 0.4%. The full safety and security report will be available in the coming days, according to Google.

What Is Still Unknown

A few things remain unproven. Every benchmark figure in this article is vendor-reported, and independent evaluations have not yet published their own Hydrogen results. The 51-versus-50 edge over Argon on the Artificial Analysis index is real but thin, and it is measured at Hydrogen’s default high setting. Google has not published a knowledge cutoff for Hydrogen on its model card or in the migration guide, so it is not yet clear how current its training data is.

The rollout is also gradual and uneven. Free users get Hydrogen only through automatic selection, not by choosing it, and Google has not given dates for when each product surface gets it. The Carbon and Argon price step on 1 December, when Carbon’s rates double, will also change the value calculation for teams choosing between tiers. And the safety report Google promises in the coming days has not yet appeared, so the safety figures above are Google’s own claims until independent testing confirms them.

Benchmark figures in this article are as reported by Google and have not been independently verified. Pricing is Google’s published API pricing as of 12 November 2026. Forward-looking statements, including rollout timing, are Google’s plans and can change.

Frequently Asked Questions

What is Gemini 4 Hydrogen?

Gemini 4 Hydrogen is Google’s fast, lower-cost AI model, launched on 12 November 2026. It is the third member of the Gemini 4 family, alongside the frontier models Carbon and Argon, and it replaces Gemini 3.8 Flash as Google’s fast tier. It is built for everyday coding and agentic workflows: writing code, running tests, verifying changes, doing research and carrying multi-step tasks to completion.

How much faster is Gemini 4 Hydrogen than Gemini 3.8 Flash?

Google says Hydrogen finishes tasks 2.5 times faster than Gemini 3.8 Flash at its default high thinking setting. On Terminal-Bench 4.1, it scores 57% against Flash’s 18.2%, more than tripling the result. In early partner testing on complex graphs, SQL and data-heavy enterprise workflows, it ran 2 times to 2.8 times faster while using 12% to 23% fewer tokens than Flash.

How much does Gemini 4 Hydrogen cost?

Hydrogen costs $1.50 per million input tokens and $7.50 per million output tokens through the Gemini API and Google AI Studio, flat across its full 1 million-token context window. For consumers it is free in Google Search and the Gemini app. It is the cheapest of the three Gemini 4 models and the only one with a free tier.

Can free users use Gemini 4 Hydrogen?

Yes, but not by choosing it directly. In the Gemini app, free users get Hydrogen automatically when Google’s system decides it suits the prompt. Paid users can pick it from the model picker, where it is labelled “Everyday help”. It is also rolling out in Google Search for all users worldwide.

What are the four thinking levels?

Hydrogen has four thinking levels: low, medium, high and xhigh. High is the default and xhigh is new. Higher levels give deeper reasoning at the cost of more time and tokens, so you only pay for the depth a task requires. At xhigh, Hydrogen reaches 54 on the Artificial Analysis Intelligence Index, against 51 at the high default.

Is Gemini 4 Hydrogen better than Gemini 4 Carbon and Argon?

Not at reasoning. Carbon and Argon are the frontier models for deep, long-horizon work, and Hydrogen is the fast execution tier. Hydrogen does edge Argon on one benchmark, scoring 51 against 50 on the Artificial Analysis Intelligence Index at its default setting, but that gap is thin. Hydrogen’s advantage is speed, price and availability: it is 2.5 times faster than Flash, costs $1.50 per million input tokens, and is the only Gemini 4 model free users can reach.

About the Author

Becky Simms

Becky Simms

Becky Simms is a technology writer with 9+ years in consumer electronics, streaming services, and device troubleshooting. Becky verifies tech news and outages against official status pages and verified outage reports, covering service outages, AI products, tech companies, and new technology coverage, and separates what is confirmed from what is still unproven. Becky’s guides show readers how to check a service themselves, right now, instead of relying on rumours.