Google Releases Gemini 4 Hydrogen: Price, Speed, Benchmarks

Google’s Gemini 4 Hydrogen costs $1.50 per million input tokens and $7.50 per million output. Here is what Google claims and what third parties add.
What Happened?
On November 12, 2026, Google released Gemini 4 Hydrogen, a new model in its Gemini 4 line. Google says it is rolling out globally across developer, enterprise and consumer products. The model is priced at $1.50 per million input tokens and $7.50 per million output tokens, and it supports a context window of 1 million tokens.
Sources: Google, “Introducing Gemini 4 Hydrogen for Agentic Workflows”
What Does the Source Say?
Google’s announcement confirms the core pricing and says the rate applies “flat across the full context window.” It also says output can run up to 1 million tokens. Developers can access Hydrogen through the Gemini API, Google AI Studio and Android Studio. Enterprises get it through Gemini for business, Google Antigravity and Gemini in Google Sheets for eligible subscribers. In the Gemini app, paid tiers see it in the model dropdown, while free users may get it through automatic model selection when a prompt suits it.
The post makes performance claims that come from Google and its partners. It says Hydrogen completes tasks 2.5 times as fast as Gemini 3.8 Flash at its default setting, and it reports 57 percent on Terminal-Bench 4.1, against 18.2 percent for Flash. It also gives an Artificial Analysis Intelligence Index score of 51 at default high thinking, versus 50 for Gemini 4 Argon. Google does not list explicit weaknesses in the post.
Sources: Google, “Introducing Gemini 4 Hydrogen for Agentic Workflows”
What Is Still Unknown?
Most performance figures are Google’s own. Independent confirmation so far is limited to the Artificial Analysis score, which VentureBeat reports as 51 at high reasoning, against 31 for Flash at the same setting. I could not open Artificial Analysis’s page directly, so I cannot confirm the ranking that some coverage describes as “near the top” of its index. The gap between Hydrogen and Gemini 4 Argon on that index is one point, which is small.
The post itself also has inconsistencies. It cites both Terminal-Bench 4.0 and 4.1, and some image captions refer to “Gemini 3.5 Flash” while the text says “3.8 Flash.” Google’s Terminal-Bench table also shows Claude Sonnet 5.5 at 63.9 percent, ahead of Hydrogen’s 57 percent, per VentureBeat’s review of that table. Partner results, such as Box’s accuracy rising from 74 to 79 percent, come from its chief executive’s post on X and are not independently audited.
Sources: VentureBeat, “Google launches Gemini 4 Hydrogen, a faster, smarter, pricier Flash successor” · Google, “Introducing Gemini 4 Hydrogen for Agentic Workflows”

Why Does It Matter?
For developers, the per-token price is only part of the cost. VentureBeat reports that Flash’s current rate of $0.75 input and $3.75 output is an introductory price through December 31, 2026. From January 1, 2027, Flash’s scheduled standard rate matches Hydrogen’s. So Hydrogen is more expensive than Flash today, but not against Flash’s scheduled rate.
Cost per completed task may matter more. Using Artificial Analysis figures as VentureBeat reports them, Hydrogen costs about $1.87 per benchmark task at high reasoning, against $7.77 for Flash. That is a benchmark-based estimate, roughly 76 percent lower: (7.77 − 1.87) ÷ 7.77 ≈ 0.76. It reflects Artificial Analysis’s test tasks, not any particular production workload.
Sources: VentureBeat, “Google launches Gemini 4 Hydrogen”
Our Analysis
Hydrogen looks best suited to agentic and coding workloads, not deep reasoning. Google’s own framing points the same way: it describes frontier models as built for “deep, long-horizon reasoning” and Hydrogen for “rapid execution across diverse agentic workflows.”
The trade-off sits in the reasoning settings. VentureBeat reports that moving from high to xhigh reasoning did not change Hydrogen’s Terminal-Bench 4.0 score, which stayed at 54 percent, but raised the estimated cost per task from $1.87 to $3.16. The same report says the new API drops the older thinking_budget parameter and the MINIMAL setting, so migrations need code changes. Our reading is that the advantage depends on the workload, and developers should test it on their own tasks before switching.
Sources: Google, “Introducing Gemini 4 Hydrogen for Agentic Workflows” · VentureBeat, “Google launches Gemini 4 Hydrogen”
Sources and References
1. Introducing Gemini 4 Hydrogen for Agentic Workflows. Google. November 12, 2026. Primary. https://blog.google/innovation-and-ai/models-and-research/gemini-models/introducing-gemini-4-hydrogen/. Supports the release, the announced rates, the context window, availability, the 2.5x comparison, and Google’s positioning.
2.Gemini 4 Hydrogen model card. Google DeepMind. Published November 12, 2026. Primary. https://deepmind.google/models/model-cards/gemini-4-hydrogen/. Supports the benchmark table, the context window, the limitations, the knowledge cutoff and the refusal regression. It does not list prices.
Saumya Dawande
B.Tech AIML @ oriental institute of science technology bhopal
Engineering and tech journalist. I love exploring the impact of emerging technologies on global defense, sovereignty, and everyday life. Always looking for the real story behind the headlines.



