Everyone waited for Gemini 4 Hydrogen. Google shipped Gemini Decision 1 instead.
$0.10/1M input, 256K context, native image/audio/video/PDF.
Not a chatbot, a router. Exactly what my agent loops need for cheap yes/no calls.
@OfficialLoganK Gemini 4 Carbon in Antigravity is great. When does Flash get the same tool-calling upgrades? Most of my agent loops don't need the big model.
Google shipped a swarm of agents before shipping the capacity to run one.
Antigravity /swarm, day one: "Model unavailable, retrying in 46s (attempt 5/9)" 🫠
Honestly, it feels disrespectful to paying users.
I pay for Google AI Pro every month, and Gemini 4 Carbon still sits behind the $99.99 Ultra paywall. Not even a small quota for Pro.
Pro users are paying customers too, not free-tier traffic. Give us access to the new model, @Google@GoogleAI.
Gemini 4 Carbon is out, and as a dev I can't wait to throw real projects at it in Antigravity 🔥
Only catch: in Antigravity it's Ultra-only for now. I'm on AI Pro, so I'm still building with Argon (which is solid), but Carbon is the one I want for coding.
@GoogleAI any plans to open Carbon in Antigravity to Pro? Even a small quota would be great 🙏
Used Antigravity Live Mode for an hour and I'm loving it. One problem: every command still runs sequentially through the main agent. Live Mode should hand the small tasks to subagents and run them in parallel. That's the actual unlock.
@kevinhou22 I think Live Mode should delegate smaller tasks to subagents and run them in parallel instead of routing everything through the main agent. I noticed every command runs sequentially through the main agent.
Antigravity now has Voice mode powered by Gemini Live. It's been a game changer internally on how many people work with a lot of folks now building hands-free with headphones in the office. Hope you all enjoy!
As always, more to come today 👀
Your Harness. You Define It.
Meet Creator Mode in DeepSeek Harness.
Turn ideas into plugins that you can use, from a design workspace to a sandbox game or a 3D LLM visualizer.
What will you create?
@yazins I am making a desktop app that helps developers and test teams manage and automate Android devices. It brings device controls into one app and can also be operated through the command line or AI agents via CLi/MCP/GUI