My team in GDM Frontier AI is hiring (Mountain View). If you're a researcher interested full duplex modeling, multimodal LLMs (Gemini), modality gap, joint speech/text modeling, PGMs, streamable generative models, and representation learning for language modeling -- DM me!
@_arohan_ People still play go and chess... I'll still be enjoying my computer hobbies even when it's "pointless". It might even be more fun when it's absolutely out of the question that it's commercially valuable ๐
I' m so addicted to @GoogleMagenta RealTime 2 ๐น
so to justify playing with it 24/7 I ported the real-time apps to @huggingface Spaces ๐ค (and ported the entire lib to pytorch/transformers too)
Play live in real time on
https://t.co/5JSl6KCNnU
MRT2 runs on MacBooks by leveraging the MLX runtime to efficiently execute it on Apple Silicon.
The framework has been a fantastic link between Python and C++, and thanks to @awnihannun for offering guidance on the MLX integration!
Really proud to share something weโve been working on for a while: Magenta RealTime 2 (MTR2), a live music model that is highly interactive (MIDI, audio, text, lots of parameters) and low-latency (~200ms end-to-end), and runs locally on a MacBook!
Introducing Magenta RealTime 2 ๐บ
- Open model for live music generation
- Just 2.4B parameters, perfect for on-device
- Low latency control
- Control with audio, MIDI, and text
We're releasing it with a series of apps to experiment directly in Mac!
Introducing Magenta RealTime 2 (MRT2): the live music model you can play as an instrument.
MRT2 offers MIDI and prompt controls, and runs natively on a MacBook with <200ms latency.
Open weights. Open source inference engine. Suite of apps and plugins.
Hear what it can do and try it out for yourself below ๐งต
With the introduction of the TPUv8t, their new training focused TPU, Google unveiled a new scale-out network architecture called Virgo. Virgo is able to interconnect up to 134,400 chips with up to 47 Pbps of non-blocking bi-sectional bandwidth. (1/4)๐งต
We've got a new model coming out next week! We've been having a lot of fun playing with it, and I hope you will tooโฅ๏ธ
We'll be celebrating by presenting at the AI Music Summit at Berklee and helping teams at the hackathon afterwards build some wild new musical instruments ๐ธ
Say hello to Antigravity CLI ๐๐ป
๐งโ๐ป - Written in Go for a snappy feel
โจ - Available with Gemini 3.5 Flash today
๐ค - Built for async workflows and subagents
โ๏ธ - Same tools and app server as Antigravity 2.0
Get started and install it today ๐
Today, we introduced Gemini 3.5 Flash โก Our most capable coding and agentic model โ where "fast" and "best" aren't a tradeoff. Try it now across Antigravity, AI Studio, Gemini App, and AI Mode.
Welcome to Gemini 3.5 Flash, our most powerful model to date. It pushes the frontier of intelligence, speed, and cost putting 3.5 Flash in a class of its own.
We spent the last 6 months making sure Flash is great for real world use cases. It's available everywhere now!
We asked our agents to build a working operating system from scratch using @Antigravity 2.0 and Gemini 3.5 Flash.
It took:
โฑ๏ธ 12 hours
๐ค 93 parallel sub-agents
๐ 15k+ model requests
๐ง 2.6B tokens processed
๐ธ Less than $1K in API credits
To build a functioning OS from scratch.
#GoogleIO
Introducing Gemini Omni ๐ฎ........ Omni is our new model that can create anything from any input โ starting with video (think Nano Banana but for video). Available in the Gemini App, Flow, and YouTube, with API support coming soon!