Deedy Sholto Douglas As swyx 📍 @aiDotEngineer’s Pareto frontier graph shows, Gemini is arguably the real story of the past few months

everyone comparing deepseek-r1 to o1 and forgetting about Gemini 2 Flash Thinking which is better than r1 on every cost and performance metric

The progress with our Gemini reasoning models is actually wild, we are in the GPT-2 era of scaling reasoning! The main delta is that the models are actually super useful in their current form and getting better week over week. The future is exciting...

Making LLMs run efficiently can feel scary, but scaling isn’t magic, it’s math! We wanted to demystify the “systems view” of LLMs and wrote a little textbook called “How To Scale Your Model” which we’re releasing today. 1/n

It's been a week and Gemini 2.0 Flash has already overtaken one of the Sonnet endpoints on OpenRouter for daily usage. A lot of it is coming from Coding use. Can't beat the price: performance ratio. ♊⚡📈

Our team at GDM Princeton is hiring! if you want to work on theoretically founded next-gen architectures for LLM, please apply here: sites.google.com/view/gbrainpri…

checkmate

Gemini flash 2.0 experimental is the first model I feel that really generalizes over different modalities. It also feels like real reasoning even when it gets it wrong. It seems to think outside the box in some cases.

BREAKING: Gemini 2.5 Pro is now #1 on the Arena leaderboard - the largest score jump ever (+40 pts vs Grok-3/GPT-4.5)! 🏆 Tested under codename "nebula"🌌, Gemini 2.5 Pro ranked #1🥇 across ALL categories and UNIQUELY #1 in Math, Creative Writing, Instruction Following, Longer

1/ Today we are releasing Gemini 2.5 Pro Experimental, our newest Gemini model with integrated “thinking” and significant performance gains. Very proud of the whole team! 🧵

Introducing Gemini 2.5 Pro Experimental! 🎉 Our newest Gemini model has stellar performance across math and science benchmarks. It’s an incredible model for coding and complex reasoning, and it’s #1 on the lmarena.ai leaderboard by a drastic 40 ELO margin. Only a handful of

#2 only to 2.5 Pro :) Another amazing collab across the board! A special thank you to my awesome team Arnaud Autef Arun Ahuja Geng Yan who were instrumental in getting this pretrained! So many more people I need to list here who helped across the stack---too many to tweet!

Another update - Ran Gemini 2.5 Flash (Auto Thinking and Non-Thinking). See the comparison below to other thinking models. Interesting curve for Gemini 2.5 Flash Non-Thinking! Meanwhile Gemini 2.5 Flash Thinking (Auto) matches Gemini 2.5 Pro! I'm still working on o3 access and

🚨Breaking: Google DeepMind’s latest Gemini-2.5-Pro is now ranked #1 across all LMArena leaderboards 🏆 Highlights: - #1 in all text arenas (Coding, Style Control, Creative Writing, etc) - #1 on the Vision leaderboard with a ~70 pts lead! - #1 on WebDev Arena, surpassing Claude

📢We’re excited to share that we’ve raised $100M in seed funding to support LMArena and continue our research on reliable AI. Led by a16z and UC Investments (University of California), we're proud to have the support of those that believe in both the science and the mission. We’re

There was a lot of announcements at IO, easy to overlook the new 2.5 Flash. It's pushing new boundaries in capability vs speed!

Great to see 2.5 Flash improve in their utility for both the reasoning and non-reasoning slices. It's an incredible model for most use cases. We're excited to see what you all build with it.