Google launches Gemini 4 Neon AI model with 1 million-token context and $3 pricing

Google has launched Gemini 4 Neon, a new artificial intelligence model aimed at complex software engineering, knowledge work and scientific research, the company announced on 18 November 2026. The model sits between the flagship Gemini 4 Carbon and the faster, lighter Gemini 4 Hydrogen in Google's growing Gemini 4 family.
Koray Kavukcuoglu, senior vice president at Google DeepMind and Google's chief AI architect, introduced Neon as bringing "frontier-level intelligence with fast end-to-end task completion" across coding, analytical reasoning and enterprise workflows. It is the fourth Gemini 4 model to arrive in a matter of weeks, following Carbon, Hydrogen and Argon.
Price, context window and thinking levels
Gemini 4 Neon costs $3 per million input tokens and $15 per million output tokens, with cached input priced at $0.15 per million tokens, a 95 per cent discount. It offers a unified context window of one million tokens, allowing it to work across spreadsheets, PDFs, charts, audio and video in a single session and turn them into structured deliverables.
The model has four thinking levels — low, medium, high and xhigh — so developers can trade intelligence against token spend for each workload. At its default high setting, Google says Neon delivers a 32 per cent lower cost per task than Carbon, while at xhigh it still costs at least 36 per cent less per task than competing models at their highest settings.
Benchmark results claimed by Google
Google reports that Neon ranks third overall on the Artificial Analysis Intelligence Index with a score of 61 at its xhigh thinking level, behind larger, slower models. It ranks second on the Vals Index with 70.9 per cent, and first on the Artificial Analysis Coding Agent Index with a score of 58.6.
On coding benchmarks, Google cites 65.2 per cent on Terminal-Bench 4.1 and 46.2 per cent on FrontierCode 2.0 at the default high setting. The company also says Neon tops Zapier's AutomationBench with 56.3 per cent and leads on Andon Labs' Vending-Bench 2, a long-horizon simulation of running a business. These are Google's own reported figures from independent evaluation providers, not results DailyByte has tested.
Built for long-running coding agents
Neon is engineered for long-running agent workflows, sustaining performance through complex, tool-heavy tasks whether running on its own or acting as a fast orchestrator directing parallel Hydrogen subagents in Google Antigravity, the company's agentic development platform. Google says it cuts redundant file reads, shares progress as it works and handles context compaction reliably in long sessions.
For production code, the model is designed for multi-file debugging and refactoring across repositories with targeted edits. On SWE-sweep, Google reports Neon leads on the mini-SWE harness with 12.8 per cent and reaches 39.0 per cent inside the Antigravity harness.
Cyber-defence variant and safety
A security-tuned variant, Gemini 4 Neon for cyber, is aimed at helping defenders write more secure code and find vulnerabilities before attackers do. Google says it reaches state-of-the-art results on the hard tier of AutoBaxBench at 76 per cent, and 90.5 per cent on SEC-bench Pro for finding code vulnerabilities. The cyber variant is rolling out to trusted defenders through Google's Fairwind Program over the coming days.
On safety, Google says Neon builds on the Gemini 4 Carbon safety stack, with real-time safeguards against cyber and biosecurity misuse, and records its lowest fabrication rate yet in internal Antigravity testing. The company says the model resisted every attack in Gray Swan's indirect prompt injection benchmark at its highest thinking level.
Where Gemini 4 Neon is available
Gemini 4 Neon began rolling out on 18 November in the Gemini API in Google AI Studio, in Gemini for business through the API and model dropdown, and in Google Antigravity for paid Google AI Pro and Ultra subscribers. In the Gemini app, Pro and Ultra subscribers can select Neon from the model dropdown, where it takes the "deeper analysis" slot previously held by Argon, according to TalkAndroid. Google has not said whether the model will come to free Gemini users or to AI Mode in Search. A full model card with technical detail has been published by Google DeepMind.
Frequently asked questions
What is Gemini 4 Neon?
It is Google's mid-tier Gemini 4 reasoning model, positioned between Carbon for the hardest long-horizon projects and Hydrogen for fast everyday execution.
How much does Gemini 4 Neon cost?
Google prices it at $3 per million input tokens and $15 per million output tokens, with cached input at $0.15 per million tokens.
Who can use Gemini 4 Neon?
Developers can use it through the Gemini API, businesses through Gemini for business, and paid Google AI Pro and Ultra subscribers in the Gemini app and Google Antigravity.
Sources: Google — Introducing Gemini 4 Neon; TalkAndroid — Gemini 4 Neon: Google's New AI Model for AI Pro Users.