Deployments range from consumer mobile phones to orbital satellites.
I'm very proud of what the Gemma team has accomplished
From tackling communities challenges to enabling people to innovate, from running in your phone to running in satellites in space, Gemma 4 has been downloaded over 200 million times
Onwards, so excited about what's to come
Positive users praise Gemma 4 as a daily driver for local on-device workflows while negative users question its lasting utility and call out overly restrictive safety alignments.
No Digg Deeper questions have been answered for this story yet.
Gemma 4 just hit 200M downloads in only 2.5 months!
For context, total downloads across the entire Gemma family of models were at 100M when we launched Gemma 3. The community's acceleration is incredible. Thank you to everyone building with Gemma.
Watch how developers are driving real-world impact:
@googlegemma They are great models. Some of my favorites
@googlegemma Please fix image input limited to 1000 pixels
@googlegemma Loving Gemma 4, it's my daily driver for local AI powered workflows. 🚀
@googlegemma Downloaded, used for about 20 minutes - COOL. Then never touched again.
@googlegemma Gemma4-12b is amazing. It just refuses to believe it's already 2026. But 👌 otherwise
@googlegemma 200M downloads in 2.5 months? Gemma is moving faster than my motivation on a Monday morning! 🚀📈
Built a Siri AI-like experience with Gemma 4 E2B.
Just talk naturally, and it can use your phone to get things done.
Health data. Translation. Calendar. Contacts. Reminders. And more.
Supports English, Chinese, and Japanese.
Everything runs entirely on-device.
• No cloud inference • No per-request fees • No token limits This is LiveLand, the AI agent inside PhoneClaw.
@osanseviero Love to hear it, and looking forward to whatever is ahead. Early access welcome. ;)
@osanseviero 👏🏻
@googlegemma Your gemma4 models are too brainwashed on safety. It affects ALL prompts, even those that shouldn't provoke any safety trigger whatsoever. Eg medical topic -> safety trigger, talking about personality -> safety trigger, talking about user preferences in life ->safety trigger. wtf
@googlegemma @_akhaliq Using it as my main local ai model, you guys killed the with the 31B it-qat version. Can't wait for Gemma5🤓
Gemma 4 is my main go-to for local models!
Gemma 4 E2B for my iPhone 15 Pro Max, I have an agentic app I made with another model to do lil things. Gemma 4 E2B/E4B dependent on task for my MacBook 2020 16gb Gemma 4 E4B or Qwen models on the RTX 3090Ti 24gb.
You guys have a very nice model
@googlegemma what are people doign with these little models
@ahmedsaoudi @googlegemma You need to inject a system prompt
@osanseviero Really looking forward to more releases
@brinkzapp @googlegemma Sounds like a tooling issue, not a model one. Check out the model card on the subject: https://huggingface.co/docs/transformers/en/model_doc/gemma4
And you can try skills with this: https://developers.googleblog.com/bring-state-of-the-art-agentic-skills-to-the-edge-with-gemma-4/
@osanseviero Let Gemma 4 120b MoE come🙏
@googlegemma Add skills etc
@googlegemma Curious how much of this growth comes from developers versus enterprise adoption
