Google Gemma (@googlegemma) header
Google Gemma (@googlegemma) profile picture

Google Gemma

@googlegemma

The official home of Google's Gemma. Lightweight, state-of-the-art open models by Google DeepMind, built on Gemini tech. What will you build? 🚀💻

luma.com/gemma-events
Joined April 2025
  • Tweets314
  • Following0
  • Followers11K
139Media
Photo by Google Gemma (@googlegemma)Photo by Google Gemma (@googlegemma)Photo by Google Gemma (@googlegemma)Photo by Google Gemma (@googlegemma)Photo by Google Gemma (@googlegemma)Photo by Google Gemma (@googlegemma)Photo by Google Gemma (@googlegemma)Photo by Google Gemma (@googlegemma)Photo by Google Gemma (@googlegemma)Photo by Google Gemma (@googlegemma)
EmbeddingGemma 2 maps text, images, video, audio, and code, running 100% locally on your device. At just 740M parameters, you can build powerful multimodal apps running entirely in your browser via WebGPU & Transformers.js. From multimodal search to cross-modal classification and clustering, local embeddings unlock endless possibilities.
18
58
14
753
29,132
Credit to @xenovacom for building this demo nitter.cf/xenovacom/status/21075…
Xenova (@xenovacom) profile pictureXenova
@xenovacomOct 6
NEW: EmbeddingGemma 2 just released on Hugging Face. Project text, images, video, and audio into a single, unified embedding space. At just 740M parameters, it can even run 100% locally in your browser on WebGPU with 🤗 Transformers.js! Try it yourself 👇
1
14
2,541
People are building some seriously cool things with EmbeddingGemma 2 that it’s becoming hard to keep up with! The community is taking its multimodal capabilities and running with them, building everything from in-browser WebGPU photo search to native Apple Silicon acceleration. So, here’s a round up of some amazing projects 🧵👇:
29
83
10
1,167
61,629
more replies
Native Core ML & Apple Neural Engine Acceleration This Core ML conversion brings EmbeddingGemma 2 natively to macOS and iOS for maximum efficiency. Using fixed-shape functions, the text encoder runs 100% on the Apple Neural Engine (ANE) without CPU fallback, while the audio and vision encoders run hardware-accelerated on the GPU, complete with Matryoshka dimension truncation (768 down to 128). Built by @Alex_tra_memory (Fluid Inference): nitter.cf/Alex_tra_memory/status… Hugging Face: huggingface.co/FluidInferenc…
Alex (@Alex_tra_memory) profile pictureAlex
@Alex_tra_memoryOct 9
We got EmbeddingGemma 2 running 100% on the Apple Neural Engine 🍎 M5 Pro 10,080 posts sorted into topics live, ~620/s on a MacBook, then any topic splits into subtopics in under a second. Fully on-device. 🤗 huggingface.co/FluidInferenc… 💻 github.com/FluidInference/Fl…
2
2
51
6,663
DigUp: Local Multimodal File Search for macOS Think Spotlight, but it understands the context meaning inside your files. Built natively in Swift with llama.cpp and Metal, DigUp uses EmbeddingGemma 2 locally to search across photos, PDFs, code, and audio/video. Even better: it jumps straight to the exact video timestamp, PDF page, or line of code. Built by @arahim: nitter.cf/_arahim_/status/210864… GitHub: github.com/ARahim3/DigUp
Abdur Rahim (@_ARahim_) profile pictureAbdur Rahim
@_ARahim_Oct 9
Search your Mac by what's in your files, not what they're called. DigUp runs @GoogleDeepMind's new EmbeddingGemma 2 on-device: describe a video moment, a podcast minute, a PDF page, a photo or code, in any language, and dig it up. github.com/ARahim3/DigUp Free and open source. @googlegemma
1
1
1
48
4,999
Speeding up Gemma! Gemini 4 Argon enabled a 2.85x speedup on an A100 via vLLM, jumping from ~37,900 to 108,000+ tokens/sec on gemma-4-E2B-it. ✅ 100% token parity ✅ Zero quality drop ✅ Done in days, not months See the exact changes across 8 PRs: goo.gle/4zNtANl
87
203
51
2,718
215,853
Credit to @lucianommartins for implementing these optimizations!
2
3
1
68
12,439
The community has already been fine-tuning EmbeddingGemma 2 and getting great results! With just 3 hours of training on 3,000 hours of Vietnamese audio, it tops 8 Vietnamese ASR benchmarks. All without needing a heavy autoregressive decoder.
Image from Google Gemma (@googlegemma)
23
52
831
37,116
Check out the breakdown from @stablequan: nitter.cf/stablequan/status/2110…
Quan Nguyen (@stablequan) profile pictureQuan Nguyen
@stablequanOct 14
Ummm guys, I took @googlegemma Embedding 2 model and train it on 3,000hrs of Vietnamese audio. IT REACHED SOTA. All it took is 2 hours on 8xH100
Image from Quan Nguyen (@stablequan)
1
24
4,915
Can you use EmbeddingGemma 2 for Speech-to-Text? This experiment uses EmbeddingGemma 2’s unified audio-text vector space to match speech against text via vector similarity. The result? Transcribing an hour of audio in about 13 seconds, locally!
56
157
21
1,886
85,126
We love seeing the community find creative new ways to push multimodal embeddings! Check out the demo and breakdown from @TslShahir: nitter.cf/TslShahir/status/21096…
Shahir (@TslShahir) profile pictureShahir
@TslShahirOct 12
I turned @googlegemma EmbeddingGemma 2, an embedding model, into a speech-to-text model, and it turned out to be one of the fastest STT models you can run on device. On a base M3 MacBook Air it transcribes an hour of audio in about 13 seconds: 288× real time. The idea came in March with Gemini Embedding 2. It puts text, images, audio and video into one shared space, so a spoken word lands right next to the written word. If that's true, you should be able to read a transcript straight out of the embedding model. I couldnt tried it then, but last week EmbeddingGemma 2 came out as an open model, so I tried it.
1
2
46
7,325
Taking meeting notes is easy; connecting them to your local files in real time is hard. Meet Google AI Edge Foresight, an on-device macOS meeting companion powered by EmbeddingGemma 2. It auto-enhances shorthand notes and searches across local docs, images, and transcripts completely offline.
33
49
8
809
45,797
Foresight’s Live Assistant automatically detects questions in real time and pulls answers directly from your local files and diagrams - no manual searching needed. Try it on macOS: developers.google.com/edge/f…
1
2
32
4,929
Built with LiteRT, Foresight shows what you can build on-device with EmbeddingGemma 2 across mobile, desktop, IoT, and web: - Instant photo & video search - Private local RAG - Sub-millisecond intent routing - Read the technical deep dive: developers.googleblog.com/go…
1
2
29
3,188
Hardware Partner Spotlight: NVIDIA joins The Google Gemma Hardware Hackathon @agihouse_org! Build with Gemma. Run it locally on NVIDIA Jetson. Bring your next edge or physical AI idea to life. On Oct 24, join 100 selected builders, the Google DeepMind Gemma team, and hardware partners for a full day of hands-on building, from camera assistants to robotics applications. 🤖 On-site NVIDIA hardware available to borrow (availability is limited): • 15 Jetson Orin Nano Super dev kits (with NVMe SSD) • 15 Jetson Thor dev kits 🏆 The Challenge & Prizes: Every project must use a Gemma-based model and run entirely on-device. Compete for a share of $45,000 in cash prizes, plus upcoming NVIDIA hardware awards.. Bring your laptop, ideas, and even your robot! 📅 Saturday, October 2410 AM–5 PM PT 📍 AGI House · Hillsborough, CA 🔗 Apply to build: agihouse.org/events/the-goog…
Image from Google Gemma (@googlegemma)
6
20
168
12,448
Catch us at other events this week: 🎤 Open Weight Debate Night (Oct 19): luma.com/592fcwx2 🏆 Open Source AI Awards Night (Oct 23): luma.com/7ebvz0oq
Image from Google Gemma (@googlegemma)
Image from Google Gemma (@googlegemma)
1
9
3,238
Welcome to the new Gemma documentation! We've built a brand new home for developers to explore, experiment, and build. What’s new? 🌐 Global model selector for customized code snippets 🤖 Better agent support (turn any page into Markdown!) 📊 Data-driven backend for a consistent experience ⚡️ On-device demos running directly in your browser! Check it out: gemma.ai/
29
126
11
1,327
57,419
Load more