In addition to the Gemma 4, Google today also detailed the Gemini Nano 4 for Android.
The Gemma 4 is “the foundation of the next generation Gemini Nano.” Gemini Nano 4 is currently available in early access through the AICore Developer Preview. Google previously called the models “nano-v2” and “nano-v3” (found in the Pixel 10 and Galaxy S26). There are two versions:
- Gemini Nano 4 Fast [Preview, TPU]: Our best performing model based on Gemma 4 E2B, designed for fast responses.
- E2B: Optimized for high speed (3x faster than E4B model!) and low latency.
- Gemini Nano 4 Full [Preview, TPU]: Our most capable model based on Gemma 4 E4B for high quality responses.
- E4B: Designed for high thinking power and complex tasks.
On Android, Google highlights how Gemma 4 is “up to 4x faster than previous versions and uses 60% less battery. It offers “industry-leading performance with multi-channel understanding” across text, image, and audio, while natively supporting 140+ languages.
Use cases include:
Consultation: Chain of command commands and conditional statements can now be expected to return high-quality results.
- For example: “Determine whether the following discussion thread comment exceeds the community guidelines. A comment does not exceed the community guideline if it contains one or more of these reasons_for_flag: profanity, abusive language, hate speech”. If the update exceeds the community guidelines, return {true}. Otherwise, return {false, reason_for_flag}.”
Statistics: With better math skills, the model can now answer questions more accurately.
- Example: “If I receive 26 paychecks a year, how much should I contribute per paycheck to reach my $10,000 savings goal during the year?”
Understanding time: The model is now more time-sensitive, making it more accurate for use cases involving calendars, reminders, and alarms.
- Example: “The event is at 6PM on August 18th, and the reminder should be sent 10 hours before the event. Return the time and date the reminder should be sent.”
Understanding the image: Use situations involving OCR (Optical Character Recognition) – such as chart recognition, visual data extraction, and handwriting recognition – will now return more accurate results.
The Gemini Nano 4 will launch in “the best new Android devices later this year.” Any “code you write today for Gemma 4 will automatically run on Gemini Nano 4-enabled devices.”
You can sign up for the AICore developer preview here. Google will “make several updates during the preview period, including support for calling, structured logouts, system prompts, and a think mode in the Prompt API.”


FTC: We use auto affiliate links to earn income. More.

