I'm at Google's big research event today, where they're showing off how their AI can be used to help science. The tone is just miles away from any Anthropic or OpenAI event. Unspoken assumption is that humans will still be driving research decisions and questions a decade from now.
I work at Google DeepMind and yes I have tried the model known externally as “Argon”.
I can say this:
I don’t reach for Opus anymore for any work (like I did when 3.8 Flash was our top model).
I hope that it will soon be generally available for everyone to enjoy!
wait. did i read that right?
gemini 4 argon has a 1M token output limit.
let me repeat. not a 1M context window. a 1M token output limit.
that means, it can write up to a million tokens in a single response.
i got confused too at first. for context, most frontier models cap output somewhere around 128k.
that's madness.
Very excited to announce Gemini 4 Argon, our new Frontier AI model, and a huge step forward across key capabilities. We’re focused on rolling it out responsibly starting with government and trusted cyber defenders through our Fairwind Program today, before wider availability soon
Introducing Gemini 4 Argon, our new frontier model, rolling out to cyber defenders starting today, and more widely as soon as possible. I am really excited by the progress we have made here. Argon is priced at $2 in and $10 out during introductory pricing!
Anyone who has ever demoed AI live knows this: non-deterministic products + wifi + unpredictable conditions can make it super tough to handle.
You can go 2 routes: polished pre-recordings or taking the risk of going live.
Hats off for choosing the 2nd option 🫡
Indeed! My customers use multiple AI models but when it comes to image/video their #1 choice is often NanoBanana/Veo.
Happy we’re seriously leaning on coding now. Onwards!! ✨
🔥 Logan on Gemini 4.0: Resources should have been poured into Coding much earlier instead of prioritizing Nano Banana. Beta testing has already started.
Looking back, it's crystal clear that Coding should have received an order of magnitude more investment much sooner. It wasn't that its importance was unknown at the time; the priority was simply diverted to image models.
Nano Banana was definitely a massive hit, but the real bottleneck for developers has always been writing code and Agents.
That "3.8 Flash" that suddenly surged in the Arena is absolutely crushing coding and long-context tasks—don't tell me that's still just the standard Flash.
Google is finally aiming in the exact right direction this time!
#Gemini4 #Gemini4Pro #GoogleAI #Google #LoganKilpatrick
At #dotAI in Paris today, and it's always refreshing to see how different companies are building and managing AI #agents.
Great talks, I'll be sharing takeaways starting next week.
One of the best parts of events like this: seeing what others are doing, in ways different from your own. It doesn't make you question your choices it just fires a new neuron in your brain.