Google has begun internal testing of a model called “Hydrogen.” It is a lightweight model that is faster and cheaper than existing high-performance models and is expected to become a Flash-tier product in the recently unveiled “Gemini 4” series. As Google employees’ internal testing is confirmed and related performance evaluations are being shared among external developers, there are indications that the launch is approaching.
Business Insider reported on the 11th (local time) that Google recently granted some employees access to a new model known by the codename Hydrogen.
According to the report, one Google engineer described the model as “excellent in performance, fast, and efficient” in an internal message this week. Employees are testing the model on Google’s internal AI development platform, “Jetski,” which allows employees to use new AI models before they are released to the public.
One employee participating in the testing described Hydrogen as being similar in nature to the existing Gemini Flash series. Flash is a lightweight model lineup that surpasses the high-performance “Argon” model in speed and cost efficiency.
Test results for a model believed to be Hydrogen are also being shared in external developer communities. Some users evaluated it as producing results close to those of the high-performance Argon model while being much faster than the existing “Gemini 3.8 Flash.” In one comparison, the existing model took about 11 minutes to complete the same motion graphics generation task, while the new model finished in about two minutes.
However, it has not been confirmed whether the model used in external testing is actually Hydrogen.
Gemini 4 Hydrogen vs Gemini 3.8 flash - Motion Graphics Comparison
I just tested both models with a motion graphics generation prompt, and the speed gap is wild:
For context on the setup, I captured Gemini 4 Hydrogen in Battle mode, going head-to-head against 3.8 Flash, and also… pic.twitter.com/G6SpT9yzNC
— Claudia Galeán (,) (@deweytn) November 6, 2026
TestingCatalog also raised the possibility that Hydrogen could launch on the 11th. However, Google has not yet revealed the release schedule, official model specifications, or pricing. Whether it will be provided for free and the method of API support also remain unconfirmed.
This model is expected to expand the lineup of the Gemini 4 product family. Google previously unveiled the high-performance Argon model and “Carbon,” which specializes in cybersecurity and long-running tasks. If Hydrogen is also released, the range of model choices is expected to broaden from high-performance reasoning to low-cost, high-speed processing.
Reporter Park Chan cpark@aitimes.com
