Google shipped a new AI model Wednesday. Its name is Gemini 4 Neon. Google DeepMind posted the news on X.
Neon sits between two other Gemini models. Carbon handles the hardest, slowest jobs. Hydrogen handles quick, simple tasks. Neon splits the difference.
The post pulled 3,578 likes within hours. It drew 139 replies too. Developers on Hacker News piled on fast.
The launch thread gathered 297 points on Hacker News. That same thread drew 288 comments. Readers argued over whether Neon beats its rivals.
Neon charges $15 for every million words of output it writes. A token is roughly a few letters or a short word. The price sits below Google's priciest model, Carbon.
Google says Neon ranks near the top of a widely watched AI scoring test. That test is called the Artificial Analysis Intelligence Index. Outside testers rank Neon just ahead of Carbon on that same index.
The model reads spreadsheets, PDFs, audio and video in one pass. It can answer a question about all of them together. Google says Neon finishes multi-step jobs faster than its bigger sibling models.
One example task is a bank statement. Neon can scan the pages. It then explains why a bill went up, in one answer.
Developers can reach Neon three ways right away. They can call it through Google's own API. They can also use it inside Google Antigravity or the regular Gemini app.
This is Google's fourth new model in about seven weeks. Carbon, Hydrogen and now Neon all shipped close together. The pace outstrips most AI labs by a wide margin.
The release says something plain about the model race right now. Google can ship a new frontier model almost weekly. Rivals now face a faster clock than they set for themselves.