Google Launches Gemini 4 Hydrogen: Faster Execution Model for Agentic Workflows at $1.50 per Million Tokens
Hydrogen replaces 3.8 Flash with 2.5x speed, strong agentic benchmarks, and a 1M-token context at flat pricing across Google products and the API.
Published
Last updated / Fact-checked on
Illustrative photo of a server-room aisle; it is not Google, Gemini, or any data center named here. Photo: Schäferle / Pixabay, Pixabay Content License.
Google introduced Gemini 4 Hydrogen on November 12, 2026, as the fast, execution-focused member of the Gemini 4 family. Designed for rapid agentic workflows, coding agents, and everyday tasks, Hydrogen brings frontier-level intelligence at speed to a broad range of Google products and developers.
Unlike the deeper reasoning models Gemini 4 Carbon and Gemini 4 Argon, Hydrogen prioritizes execution speed while maintaining high capability. Google positions it as the daily driver for agent loops, replacing Gemini 3.8 Flash across AI Studio, the Gemini API, the Gemini app, Google Search, and for paid Pro and Ultra subscribers.
Key Specifications and Pricing
Hydrogen features a unified 1,048,576-token context window that covers both input and output tokens. Pricing is flat at $1.50 per million input tokens and $7.50 per million output tokens across the entire window—unchanged from the prior Flash model in many comparisons. It supports four thinking levels: low, medium, high (default), and a new xhigh tier. Sampling parameters like temperature are ignored, and it uses implicit context caching.
Availability rolled out quickly to AI Studio (all tiers), the Gemini API, Gemini for business, the Gemini app, Google Search, and Google Antigravity. Free and Plus users may still see older models in some cases, while Pro and Ultra get full access.
Performance Claims
Google reports that at the default high thinking level, Hydrogen finishes tasks 2.5 times faster than Gemini 3.8 Flash. On the Artificial Analysis Intelligence Index, it scores 51 at high (surpassing Argon's 50) and reaches 54 at xhigh. Terminal-Bench 4.1 scores more than triple previous Flash results (57% vs. 18.2%). It cuts fabricated claims by more than 5x and handles tool-call failures or empty searches cleanly.
In agentic benchmarks, Hydrogen ranks highly on Agent Arena, leads its class on HLE-Diamond (58.8%), tops Vals Tax Agent Bench (52.3%), and ties for first on Vals Legal Research Bench (55.8%). Partners reported 2x–2.8x speed gains with 12–23% lower token usage versus 3.8 Flash in some workflows.
At a Glance
| Aspect | Detail | Source |
|---|---|---|
| Launch | Nov 12, 2026 | |
| Context | 1M unified tokens | |
| Pricing | $1.50 in / $7.50 out per MTok | |
| Speed | 2.5x vs 3.8 Flash (high) | |
| AA Index | 51 (high), 54 (xhigh) | Google / AA |
| Terminal-Bench 4.1 | 57% |
Implications for Developers and Agents
Hydrogen is particularly suited for iterative coding, debugging, deep research, data analysis, and multi-step agent tasks. Its speed and reliability in tool use make it attractive for production agent systems where latency and cost matter. Google notes it adapts well when tools fail, reducing error loops common in earlier models.
This launch continues Google's elemental naming (Carbon, Argon, Hydrogen) and expands Gemini 4 access beyond frontier tiers. Independent evaluations will be key to confirming the claimed speed and quality gains.
What to Watch
Real-world agent performance, independent benchmark results, adoption in enterprise workflows, and any updates to thinking levels or caching. Comparisons with concurrent releases from OpenAI and Anthropic will shape developer choices.
How We Reported This
Based on Google's November 12 announcement blog, model documentation, and coverage from Ars Technica, 9to5Google, MarkTechPost, and The Decoder, reviewed November 13. All performance, speed, and benchmark figures are as reported by Google. No hands-on testing was conducted.
Published and fact-checked November 13, 2026.
Sources
- Introducing Gemini 4 Hydrogen, Google Blog, November 12, 2026, https://blog.google/innovation-and-ai/models-and-research/gemini-models/introducing-gemini-4-hydrogen/
- Google Cloud documentation on Gemini 4 Hydrogen
- Ars Technica, 9to5Google, MarkTechPost, The Decoder coverage, November 12–13, 2026
- Artificial Analysis Intelligence Index references in coverage
DailyMech used AI tools to help research and draft this article; it was fact-checked by our editors.