Google releases Gemini 4 Hydrogen, a fast model that replaces 3.8 Flash and upgrades Search

What Google released and where it is available

Google introduced Gemini 4 Hydrogen on 12 November 2026. The company positions it as the fast-execution option in the Gemini 4 line. Its frontier models, Gemini 4 Carbon and Gemini 4 Argon, are meant for deep reasoning on hard problems, while Hydrogen targets quick work across agentic workflows, meaning tasks in which an AI carries out several steps on its own, as TechCrunch explains.

Google says the model is available now in all tiers of Google AI Studio, the Gemini API, Gemini for business, the Gemini app and Google Search. In the Antigravity development platform it is offered to paid Google AI Pro and Ultra subscribers. Ars Technica reports that Hydrogen has taken the place of Gemini 3.8 Flash across the company's products.

Ars Technica contrasts this with the earlier Argon and Carbon releases. According to the publication, those were staged, began with cybersecurity researchers, and at general availability were limited to people on the most expensive AI subscription. It describes the limits on Hydrogen as much looser, with access starting immediately.

Speed, benchmarks and pricing as Google states them

Google says Hydrogen runs 2.5 times faster than Gemini 3.8 Flash at the default high setting. It also says the model cuts fabricated claims by more than 5x. On Terminal-Bench 4.1 the company reports 57% against 18.2% for the older model. Ars Technica describes this as more than a tripling of the Flash score.

On the Artificial Analysis Intelligence Index, Google reports a score of 51 for Hydrogen at high thinking, against 50 for Argon, and 54 at the highest setting. TechCrunch and Ars Technica repeat the comparison with Argon. Ars Technica frames it as the xhigh score of 54 against Argon's 50. These are the company's figures, and the sources do not include independent replication.

Google also reports a number 5 ranking on Agent Arena, the live benchmark from arena.ai, for multi-step work such as deep research and data analysis. It lists 58.8% on HLE-Diamond, first place on Vals Tax Agent Bench at 52.3%, and a tie for first on Vals Legal Research Bench at 55.8%. It cites 53.4% on its new Multimodal Knowledge Agent benchmark and 71.6% on Agents' Last Exam v1.1. Google says partners testing complex graph, SQL and data-heavy enterprise workflows saw runs 2x to 2.8x faster, with token use down 12% to 23% versus 3.8 Flash.

Hydrogen has four thinking levels: low, medium, high (the default) and a new tier called xhigh. Google says the levels give a predictable range of quality, latency and cost. Pricing is $1.50 per million input tokens and $7.50 per million output tokens, flat across the whole 1 million-token context window. Google adds that the model accepts up to 1 million input tokens and can produce up to 1 million output tokens.

Orchestrator and multi-model workflows in Antigravity

Google presents Hydrogen as usable on its own for everyday coding work such as writing code, running tests and verifying changes. It can also be paired with Carbon and Argon. To automate that pairing, the company is introducing Orchestrator in Google Antigravity, which switches between models such as Carbon and Hydrogen within one session.

On Terminal-Bench 4.0, Google says Orchestrator reached task success rates close to Carbon's while using on average 24% less quota and finishing 19% faster. For parallel agents, Carbon can act as coordinator while Hydrogen runs worker and verifier agents in isolated workspaces. Those figures come from Google and are not independently confirmed in the reports.

What changes in Search and the Gemini app

Google calls Hydrogen the biggest model upgrade to Search so far. The Verge, TechCrunch and Google describe the same changes. Search should handle questions with several parts, hard-to-read typos and highly specific requests, and address each part of a query. Google credits better instruction following and tool calling.

The model also feeds Search's query fan-out technique, which splits a prompt into several subqueries run at once. Google says this connects users to a wider range of links, including sites they might have missed. Responses in AI Mode will include more images, structured answers, comparison tables and custom tools. Google adds that answers should improve in all languages AI Mode supports.

In the Gemini app, Google says Hydrogen follows tone, format and rules more closely across multi-turn conversations. It also pulls context from uploads, Gmail, Google Drive and connected apps such as Airtable and Dropbox, and calls tools more often when a task needs them.