Gemini is really good and we keep making it better. In quality and performance but also in cost efficiency.

With outputs costing $0.40 per million tokens and a 0.7% hallucination rate (Vectara), Gemini 2.0 Flash is best-in-class and has been my go-to model for most applications.

Excited to be speaking at GDC next month with @estragone.bsky.social! Join us at the ML Summit for "SIMA: Developing General Agents with Video Games".

A look at the History of AI research with games and the latest on SIMA, our AI agent that follows language-instructions in 3D games. #GDC2025

Gemini 2.0 Pro Experimental ranks #1 across all categories in Chatbot Arena! 🚀🥇

Gemini 2.0 Flash is top-3 in Coding, Math, and Hard Prompts. Our new model Flash-lite is top-10 across categories.

Gemini 2.0 Flash is now generally available via the Gemini API in Google Al Studio and Vertex Al 🚀

Also introducing:
🔵 2.0 Pro Experimental, excels at coding, comes with 2M tokens context
🔵 2.0 Flash-Lite, our most cost-efficient model
🔵 2.0 Flash Thinking Exp in Gemini App

Excited to introduce Genie 2: our most capable and general foundation world model. It can generate a diverse array of consistent and interactive 3D worlds from a single image, playable with keyboard and mouse inputs for up to a minute. 🧞🎮🌎

Our latest Gemini iteration is #1 on Chatbot Arena IN EVERY CATEGORY! 🥇🎉

Try Gemini-Exp-1206 for free in Google AI Studio and the Gemini API now.

Gemini was released 1 year ago! 🧁🥳

From 1.0 state-of-the-art multimodal capabilities to giant leaps forward with 1.5 and 2M tokens long context to the Project Astra research prototype...and that's only year one 😀

Congrats to everyone at Google DeepMind and our partners at Google!

I've been promoted to Al Research Group PM at Google DeepMind! It was a privilege to contribute to groundbreaking projects like Gemini, Astra, SIMA & Genie with such talented and inspiring teams.

Excited to continue this journey and contribute to our ambitious research ahead. Onwards and upwards! 🚀