Google announced Gemini 4 Hydrogen on 12 November 2026 as a fast, efficient model for everyday coding and agent workflows. Google prices it at $1.50 per million input tokens and $7.50 per million output tokens, flat across a 1-million-token context window. The announcement positions Hydrogen below Google’s frontier models, Carbon and Argon, which it says handle deep, long-horizon reasoning. The benchmark results below are Google’s own, and Lens has not checked them.

Benchmark and speed claims in Google’s announcement

  • Artificial Analysis Intelligence Index: 51 at the default “high” thinking level, against 50 for Argon, and 54 at the highest “xhigh” setting.
  • Coding and agents: 57% on Terminal-Bench 4.1, against 18.2% for Gemini 3.8 Flash; 71.6% on Agents’ Last Exam v1.1; fifth on Agent Arena; leads its class on AutomationBench-AA, with no score given.
  • Knowledge and professional tasks: 58.8% on HLE-Diamond, which Google calls the top score in its class; 52.3% on Vals Tax Agent Bench, ranked first overall; 55.8% on Vals Legal Research Bench, tied for first; 53.4% on a multimodal knowledge-agent benchmark, ahead of other models according to Google.
  • Speed: About 2.5 times faster than Gemini 3.8 Flash at the default setting. In early partner tests on graph, SQL and data-heavy enterprise workloads, Google reports 2x to 2.8x faster results with 12% to 23% fewer tokens, and more than 30% faster outputs in Google Sheets. There are four thinking levels, low, medium, high (default) and xhigh.
  • Orchestrator: Google says its Orchestrator, which alternates between Carbon and Hydrogen in one session, reached near-Carbon task success on Terminal-Bench 4.0 while using about 24% less quota and finishing 19% faster. It suggests Carbon as a coordinator, with Hydrogen running worker and verifier agents.

Where it is available

  • Developers: The Gemini API, Google AI Studio and Android Studio. Google Antigravity is available to paid AI Pro and Ultra subscribers.
  • Businesses: Gemini for business through the API or the model dropdown, plus Antigravity and Google Sheets for eligible subscriptions.
  • Consumers: The Gemini app model dropdown, where free users get it through automatic model selection when it fits the prompt, and Google Search, which Google says is rolling out to all users globally. In Google Sheets it starts with AI Pro and Ultra subscribers.
  • Not mentioned: The announcement does not mention Vertex AI.

Caveats

Most comparisons are against Gemini 3.8 Flash, the baseline Google chose, and the announcement gives no direct Hydrogen-versus-Carbon benchmark numbers. Lens did not check the Artificial Analysis, Vals or Terminal-Bench figures on those organizations’ own pages, and the announcement does not say how many runs the numbers represent.

Sources and notes

Source: Google, “Introducing Gemini 4 Hydrogen” (12 November 2026), read as extracted text. The release was also listed in The Neuron’s 13 November digest. Lens did not test the model.

Image: Exterior of Google's data center in The Dalles, Oregon, with cooling towers in front of hills, 17 May 2015. Photo: Tony Webster (CC BY 2.0), via Wikimedia Commons; cropped. Stand-in image: not a photo of the Gemini 4 Hydrogen model or of the hardware that runs it.