Google has introduced Gemini 4 Neon, a new artificial intelligence model designed to handle complex coding, enterprise workflows, scientific research and multi-step tasks while reducing the cost of AI-powered operations.
Announced on November 18, 2026, the model joins Google's Gemini 4 family, positioning itself between the more powerful Carbon model and the faster Hydrogen model. Google said Neon is intended to deliver advanced reasoning capabilities with greater efficiency for developers, businesses and consumers.
According to the company, Gemini 4 Neon ranks third overall on the Artificial Analysis Intelligence Index, scoring 61 at its highest reasoning setting. It also secured second place on the Vals Index and the Agent Arena Index, which evaluates AI agents across practical tasks.
Google said the model has been designed to improve instruction-following, adapt to user corrections during ongoing tasks and reduce fabricated responses.
The model supports a one-million-token context window, allowing users to process large volumes of information, including spreadsheets, PDFs, charts, audio and video, within a single workflow.
For developers, Neon is designed to handle software engineering tasks involving multiple files, debugging, code restructuring and automated execution. It can also coordinate parallel AI agents through Google's Antigravity development environment.
The company reported that Neon scored 65.2% on Terminal-Bench 4.1 and 46.2% on FrontierCode 2.0 at its default high reasoning setting.
Beyond software development, Google highlighted the model's performance across financial analysis, accounting, spreadsheet modelling and scientific research.
Neon ranked first on the Legal Research Bench and Tax Agent Bench, according to the evaluations cited by Google. The company also reported strong results in mathematical reasoning and scientific problem-solving benchmarks.
Pricing has been set at $3 per million input tokens and $15 per million output tokens. Cached input tokens cost $0.15 per million, representing a 95% discount compared with standard input pricing.
Google said Neon delivers a 32% lower cost per task than Gemini 4 Carbon at its default high reasoning setting, based on its benchmark comparisons.
The model also introduces four reasoning levels, allowing developers and enterprises to adjust computational effort according to task complexity and cost requirements.
Cybersecurity is another focus area, with Google reporting improvements in vulnerability detection, code security and automated patching. A specialised version of Neon will be made available to eligible cybersecurity defenders through its Fairwind Programme.
Google has also outlined safety measures addressing prompt injection, unauthorised agent actions and misuse risks.
Gemini 4 Neon is rolling out through the Gemini API, Google AI Studio, Gemini for Business and Google Antigravity. Paid Google AI Pro and Ultra subscribers can also access the model through the Gemini application.
The launch expands Google's AI portfolio as technology companies compete to deliver models capable of handling increasingly complex enterprise tasks with greater efficiency.
The performance and cost comparisons are based on benchmarks cited by Google and may not reflect results across every real-world application.
Disclaimer: This article may include information derived from interviews, press releases, public statements, research, company communications and other publicly available or third-party sources. Such material may be summarised, paraphrased or contextualised for journalistic and editorial purposes. All rights in third-party content remain with their respective owners.