Google introduces Gemini 4 Neon for complex workflows
Google has introduced Gemini 4 Neon, with API and app rollouts, a one-million-token context window and four thinking levels.
Source: Google Gemini Blog

Google announced Gemini 4 Neon on 18 November 2026, positioning the model for complex software engineering, knowledge work and science workflows. Its rollout began that day across selected Gemini products and subscriptions.
What is included in Gemini 4 Neon?
Gemini 4 Neon has a 1 million-token context window and can work with spreadsheets, PDFs, charts, audio and video in the same context. Google says it can combine these inputs and turn them into structured outputs.
The model is designed for long-running tasks that use multiple tools. It can operate on its own or direct parallel Gemini 4 Hydrogen subagents in Google Antigravity, while reporting progress and managing context during longer sessions.
Google has also added four thinking levels. These let teams adjust the model’s reasoning depth and token use according to the workload, with xhigh intended for more difficult engineering, analytical and scientific tasks.
How did the model perform in testing?
Google reported a score of 61 for Neon at the xhigh thinking level on the Artificial Analysis Intelligence Index, placing it third overall. Neon ranked second on the Vals Index. It also ranked second for steerability on the Agent Arena Index, which assesses adaptation to mid-task user corrections.
For automation work, Neon ranked first overall on Zapier’s AutomationBench with a score of 56.3%. Google also reported a score of 1336 and second place globally on the Agent Arena Index, based on more than 4,200 agent sessions.
The announcement says Neon recorded Google’s lowest fabrication rate so far in internal Google Antigravity testing. It also describes progress in following multi-step instructions. These are Google’s reported results, so businesses should assess the model against their own tasks and review its outputs.
Where is Gemini 4 Neon available?
The rollout covers the Gemini API in Google AI Studio and Gemini for business through its API and model dropdown. It also covers Google Antigravity for paid Google AI Pro and Ultra subscribers, plus the Gemini app for Pro and Ultra subscribers.
Google lists pricing at $3 per million input tokens, $0.15 per million cached input tokens and $15 per million output tokens. The announcement does not identify the currency or state whether the rollout includes Australia.
A separate cyber version is due to roll out to trusted defenders through the Fairwind Program in the coming days. Google says this version covers work such as finding, validating and patching code vulnerabilities.
Assess possible business uses
For businesses assessing the model, a cautious trial could start with a defined process involving several files, repeated analysis or multiple software tools. Possible tests include organising spreadsheet information, producing structured material from mixed file types or checking internal software tasks in a controlled setting.
This is practical guidance rather than a capability claim from Google. A trial can use approved source material, a narrow task and a person responsible for checking the result. Businesses can also compare token use across the four thinking levels instead of selecting the highest setting for every task.
Businesses considering connected AI workflows can review the Automation service page. A free audit can also help identify existing digital processes that may need closer review before new automation is introduced.
See what to fix first on your own website
A free audit of your website, Google listing and tracking, with the fixes that matter most. The first results show on screen straight away.


