Gemini 4 Hydrogen
Google's fast Gemini 4 model replaces Gemini 3.8 Flash for everyday agent, coding and multimodal work. It adds a fourth thinking level, xhigh, and lists 25% below Carbon and Argon per token.
Gemini 4 Hydrogen is available on MiniRouter as . The model page has current pricing, provider routes and API setup.
About the model
- Context: 1M tokens. Shared by input, thinking and output, per Google.
- Thinking: Low · medium · high · xhigh. High by default; it cannot be turned off.
- Best fit: Fast everyday agent work. Google keeps Carbon and Argon for deep, long-horizon reasoning.
- Knowledge cutoff: March 2026. Per Google's model card.
Google list pricing
Google's list prices, before MiniRouter's fee. Cached input is $0.075 per million tokens. The rate is the same at every prompt length, and Google lists no dated change.
| Phase | Input / 1M | Output / 1M | Blended / 1M | Effective | Source |
|---|---|---|---|---|---|
| Publisher price | $1.50 | $7.50 | $3.00 | Nov 12, 2026 – open-ended | Google ↗ |
Publisher-reported benchmarks
Results from the publisher’s announcement, grouped by unit.
| Benchmark | Reported result | Reported by | As of | Source |
|---|---|---|---|---|
| Agents' Last Exam v1.1 | 71.6% | Google DeepMind | Nov 12, 2026 | Announcement ↗ |
| AutomationBench-AA | 76.1% | Google DeepMind | Nov 12, 2026 | Announcement ↗ |
| HLE-Diamond | 58.8% | Google DeepMind | Nov 12, 2026 | Announcement ↗ |
| Terminal-Bench 4.1 | 57.0% | Google DeepMind | Nov 12, 2026 | Announcement ↗ |
| OSWorld-2.1 · offline subset | 77.6% | Google DeepMind | Nov 12, 2026 | Announcement ↗ |
| Terminal-Bench-Science 0.1 | 51.4% | Google DeepMind | Nov 12, 2026 | Announcement ↗ |
Hydrogen does not lead every row: Google's model card puts Claude Sonnet 5.5 ahead on Terminal-Bench 4.1, OSWorld-2.1 and Terminal-Bench-Science 0.1.
Against Gemini 3.8 Flash, Google reports tasks finishing 2.5 times faster at high effort and Terminal-Bench 4.1 rising from 18.2% to 57%.
Artificial Analysis scores Hydrogen 54 at xhigh and 51 at high on its Intelligence Index, against 50 for Argon and 60 for Carbon, and measures about 41 and 15 seconds to the first answer token. Publisher and third-party results are not MiniRouter route measurements.
Keep exploring
Compare published results in the model rankings and coding benchmarks, or read more model releases.
For MiniRouter product updates, visit the changelog.
Sources
- Read at Google DeepMind ↗
Gemini 4 Hydrogen model card
Published Nov 12, 2026
- Read at Google ↗
Google Gemini 4 Hydrogen announcement
Published Nov 12, 2026
- Read at Google DeepMind ↗
Gemini 4 Hydrogen evaluation methodology
Published Nov 12, 2026
- Read at Google ↗
What's new in Gemini 4 Hydrogen
Published Nov 12, 2026
- Read at Google ↗
Gemini API pricing
Published Nov 12, 2026
- Read at Vercel ↗
Vercel Gemini 4 Hydrogen catalog
Published Nov 12, 2026
Editorial reference revised 2026-11-12. Catalog status comes from the generated routing snapshot.