Google has added a new model to its Gemini lineup. Gemini 4 Neon, announced on 18 November, is now rolling out to paying subscribers of the Gemini app and to developers who build on Google's AI services.
Neon occupies the middle of Google's current range. It sits above Hydrogen, the model intended for everyday requests, and below Carbon, which Google reserves for its most demanding workloads. According to a report by Business Upturn, the company is positioning Neon around coding, research and detailed analysis.
Where Neon Fits in Google's Model Family
Google describes Neon as an attempt to balance three factors: reasoning ability, speed and the cost of getting a job done. That framing places it as a practical choice for work that needs more depth than Hydrogen offers without moving up to Carbon.
The launch matters most for Google AI Pro subscribers. Carbon is not available on the Pro plan, so Neon gives those users access to a more capable analysis model than they had before, while Ultra subscribers gain another option alongside the top tier.
How Users and Developers Can Access It
In the Gemini app, Neon is available to Google AI Pro and Ultra subscribers, who can pick it from the model menu once the rollout reaches their account. Google is also offering it to businesses through the Gemini for business API and its model selector.
Developers can reach Neon through Google's paid API, Google AI Studio and Android Studio. Paid Pro and Ultra subscribers can also use it inside Antigravity, Google's coding platform.
Pricing and Context Window
Developers pay based on usage rather than relying on an app subscription. Google lists the price at $3 per million input tokens and $15 per million output tokens, with cached input priced at $0.15 per million tokens.
Neon supports a context window of up to one million tokens, which means a single request can include very large volumes of material such as long documents or extensive codebases. A large context allows more information to be supplied, but it does not guarantee that every answer based on that information will be accurate.
Coding, Research and Multistep Tasks
Google is aiming Neon at software development, research and work involving documents, spreadsheets, charts, audio and video. In Antigravity, the model can either carry out a lengthy coding job on its own or coordinate several Hydrogen agents, each handling a separate part of the work.
Google's demonstrations included debugging problems that span multiple files and building interactive software. These examples reflect what Google tested internally and should not be read as a promise of the same results on every developer's project.
For consumers, the app launch covers writing, analysis of uploaded files and research that draws on information from the web. Google also says its Tasks feature can use Neon for jobs that involve several steps.
Benchmarks Come With Caveats
Google reports that Neon scored 61 on the Artificial Analysis Intelligence Index when run at its highest thinking setting, placing it third overall. VentureBeat reported that Artificial Analysis confirmed that ranking.
The setting used is important. Neon offers four thinking levels, and more reasoning can increase both response time and the amount of paid usage consumed. A top score at the maximum setting is not directly comparable with a rival model tested at a lower level.
Price comparisons also depend on timing. As VentureBeat and 9to5Google have noted, Carbon is currently available at introductory token prices that run until 30 November. On a per-token basis, Neon is not cheaper than that temporary Carbon offer, although Google says Neon costs less per completed task in some of its tests.
What the Launch Means for Builders and Subscribers
For developers and startups building on Gemini, Neon adds a clearly priced middle option that can be matched to workloads which do not need Google's most expensive model. Teams will still need to test it on their own tasks, since speed, accuracy and savings vary with the job and the thinking level chosen.
For app subscribers, especially those on the Pro plan, the release brings a stronger analysis model into everyday use. Google has also said a separate cyber-focused version will reach eligible Fairwind defenders in the coming days, rather than being made available to every Gemini user at once.
Source: Business Upturn





