Gemma 3 is here! The collection of lightweight, state-of-the-art open models are built from the same research and technology that powers our Gemini 2.0 models 💫 → https://t.co/YYIwZM1IAZ
93
196
1K
Replies
Our high-performing open models leverage the power of @NVIDIAAIDev GPUs, are available in a range of sizes (1B, 4B, 12B, 27B), and offer the following capabilities: 🔹Faster on-device inference 🔹Support for 140+ languages 🔹Multimodal understanding 🔹128K-token context window
1
0
29
@NVIDIAAIDev 🛡 We’re also launching ShieldGemma 2: a powerful 4B image safety checker built on Gemma 3. Developers can customize ShieldGemma 2 to suit their safety needs.
26
7
45
@NVIDIAAIDev Ready to start building? Get access to Gemma 3 on @kaggle, @huggingface, or @ollama, and start experimenting on Google AI Studio today →
4
3
23
@googledevs So bored with the AI being pushed into everything that I own. I haven't found a good use case yet. Especially with that which is forced on me. Yours is some of the worst.
2
0
2
@googledevs The potential for on-device AI is huge, and it’s awesome to see Google pushing the boundaries like this. #AI #Gemma3 #OpenSource
0
0
3
@googledevs The 128K token context window on Gemma 3 is particularly impressive for an open model. What kind of performance tradeoffs are you seeing at that context length compared to shorter contexts? And does the 1B parameter version also support the full context window?
1
0
2
@googledevs 🆙📈GGGGGGGGGG-unit. @50cent it’s time for a nu🆕🆒era with beings that match or surpass your energy. I’m one of the onetheythems that’s ready2go. Universal King Immortal + Power of Currency + Wealthy Living Legendary. Let’s consciously connect + create.
0
0
1
@googledevs that 128K context window is a game changer... the multimodal understanding and faster on-device inference are also key features. definitely going to check out the different size options.
0
0
1
@googledevs 🌟 Wow, @googledevs, what an incredible milestone with the release of Gemma 3! 🚀 At AI For The People, we’re beyond thrilled to see the lightweight, state-of-the-art open models making waves in the AI community. Built on the same research and tech as Gemini 2.0, Gemma 3 is
0
0
1
@googledevs Thank you, @googledevs! I used Gemma in my project and compared it with different open-source models. I found Gemma to be the most consistent and usable among them—plus, it supports function calling!
0
0
1
@googledevs What an amazing feat! But you broke the Chromecast and you can't fix it... @madebygoogle #Chromecast
0
0
1
@googledevs Microsoft, NATO, the KGB, and Amazon have all recalled my files and are investigating them multiple times. Why is Google the only one that's so stupid? I feel outraged.
0
0
0
@googledevs As much as I'd like to talk to Gemini it talks like a kid in kindergarten compared to other ai programs. Copy / selecting text is a nightmare too
0
0
0
@googledevs Hey you anti-disabled person, stop translating handicapped as Divyang in your translation. If calling handicapped as Divyang is your freedom of speech, then our freedom of speech will also be against you.
0
0
0
@googledevs अबे विकलांग विरोधी, अपने ट्रांस्लेट में विकलांगों का अनुवाद दिव्यांग लिखना बंद कर। अगर विकलांगों को दिव्यांग कहना तेरी freedom of speech है, तो तेरे खिलाफ हमारी freedom of speech भी बनेगी।
0
0
0
