Google Releases Gemini 4 Neon, and the Announcement Gives Not a Single Benchmark Number
Gemini 4 Neon is starting to roll out, first to paying users, across the API, enterprise, and consumer sides. Google is framing it differently this time: no benchmarks, just the speed of getting work done.
Google's official account posted that Gemini 4 Neon is starting to roll out. Sundar Pichai reposted the post.
The official positioning is: frontier-level intelligence that can quickly complete complex end-to-end tasks, covering software engineering, knowledge work, and scientific workflows.
The rollout spans several areas:
- Gemini API and Google AI Studio
- Gemini for business
- Google Antigravity, for paying Google AI Pro and Ultra subscribers
- Gemini App, also for Pro and Ultra subscribers
Antigravity is Google's AI coding tool. Gemini 4 Neon launching on it first suggests Google wants to tie model capabilities to its own developer tools.
One detail is worth noting: the entire announcement contains no benchmark numbers and no comparison tables. Google uses only two phrases, frontier-level intelligence and fast end-to-end task completion. The former is the positioning; the latter is the selling point.
The term end-to-end here refers to multi-step tasks: first understand the requirement, then break it down, then call tools, then verify. Whether a model can run through the whole thing in one go affects the real-world experience more than whether it can solve a particular test question. Put plainly, what Google is selling this time is the speed of getting the work done, not scores on individual tests.
Another detail is that the rollout starts today and is gradual, not a full open release. Paying users get it first, while free users have to wait in line. Google has been using new models as a subscription benefit for some time.
As for whether it is actually fast in practice, the announcement doesn't settle that; people will have to try it for themselves once they get access.
发布时间: 2026-11-19 06:29