Google launches Gemini 4 Neon: 1M-token context, $3 per million input tokens, available to Pro/Ultra subscribers and via API
Google has announced the rollout of Gemini 4 Neon, focused on coding, knowledge work and cybersecurity, with a 1M-token context window. The Gemini app is limited to Pro and Ultra subscribers; the cyber edition is for Fairwind members only.
News date: About 6 min read
Contents
On November 18, 2026, Google announced Gemini 4 Neon, positioned as the model in the Gemini 4 family that balances depth of reasoning with efficiency. Google says it is "rolling out today," but this is a staged rollout by channel and plan rather than general availability: the API, enterprise offering and Gemini app each have different conditions, and the cyber edition requires vetting. Two common terms first: a token is the basic unit AI models use to measure and bill text, roughly a short piece of a word; the context window is the maximum amount of content the model can "read in" and keep in mind at once, so 1M tokens means it can handle a large volume of documents in one go. For developers and businesses, the point is handling long, multi-step coding and analysis work at lower cost; most of the performance figures below are results cited by Google itself.
What did Gemini 4 Neon change, and how does it differ from other models in the family?
Google describes Neon as "bringing frontier-level intelligence with fast end-to-end task completion" across software engineering, knowledge work and science workflows. Within the Gemini 4 family, Google says Carbon handles the hardest long-horizon projects, Hydrogen handles fast everyday execution, and Neon strikes the balance between depth and efficiency.
Google also says Neon has four thinking levels (think of them as settings for how long and how deeply the model reasons), letting teams tune intelligence against token spend for each workload. At the default high level, Google says Neon pairs a high Artificial Analysis Intelligence Index score with 32% lower cost per task than Carbon. At xhigh, Google says it reaches its highest intelligence score while costing more than 36% less per task than competing models at their highest settings.
Who can use it now, and on which platforms?
| Channel | Audience | Conditions / status |
|---|---|---|
| Gemini API (Google AI Studio, Android Studio) | Developers | Described by Google as the paid tier; rolling out today |
| Gemini for business | Enterprises | Via the API or the model dropdown; rolling out today |
| Google Antigravity | Paid Google AI Pro and Ultra subscribers | Rolling out today |
| Gemini app | Google AI Pro and Ultra subscribers | Neon can be selected in the model dropdown |
| Neon for cyber | Trusted defenders in the Fairwind Program | Opening over the coming days; apply by contacting your Cloud account team |
Google's announcement does not list availability by country or region, so it is not yet clear whether Neon is available where you live. Check the official product pages to confirm.
What does it mean for everyday workflows?
For software development, Google says Neon is designed for long-running agentic workflows (agents are AI that can use tools step by step on their own to complete tasks). Google says it reduces repeated file reads and reports progress as it works. In Antigravity, it can also act as an orchestrator that directs multiple Hydrogen subagents in parallel. Benchmark figures (standardized question sets used to compare model capability) cited by Google include: 1st place on the Artificial Analysis Coding Agent Index (58.6), 65.2% on Terminal-Bench 4.1, and 46.2% on FrontierCode 2.0 at the default high setting.
Regarding FrontierCode 2.0, an article by its developer Cognition (dated 2026.10.27) says the benchmark is "far from saturated," with the top score at publication, Opus 5.5, at only about 47%; its results table lists Gemini 4 Argon (27.6%) and does not list Neon. In other words, Neon's 46.2% is Google's own claim and is not independently confirmed on the Cognition page.
In the Gemini app, Google says Neon reads spreadsheets, PDFs and photos more precisely and draws on content from Gmail and Docs. The "tasks" feature in Gemini can now also run on Neon. Examples include analyzing a bill to find out why a charge spiked, or cross-checking receipts against documents.
What is Neon for cyber, and what are its limits?
Google says Neon for cyber covers finding, validating and patching vulnerabilities. By its published figures, it scores 76% on AutoBaxBench hard, 51.26% on WebExploitBench v1.2, 90.5% on SEC-bench Pro and 70.7% on CWE-bench v2 (a 3.4% uplift over Carbon). These are likewise results cited by Google.
Access is restricted. According to the Fairwind Program page, the program prioritizes governments, critical infrastructure operators and core technology platforms, and runs background checks on applicants. Participating organizations must agree to user-level authentication and phishing-resistant multi-factor authentication. They may only give access to internal security, incident response or penetration testing teams, and may not share or resell it. Malicious uses (such as building malware) are not permitted.
What does Google say about safety and reliability?
Google says Neon maintains Carbon-level safeguards. It refuses only 1.98% of tasks on Terminal-Bench 4.0 and 0% on Terminal-Bench-Science, and records a 0.09% rate of unexpected (misaligned) actions in Google's agentic safety tests. Gray Swan's indirect prompt injection benchmark tests whether a model can be manipulated by malicious instructions hidden in web pages, emails or documents. On it, Google says Neon blocked all attacks (0.0% attack success rate) at the highest thinking level with real-time safeguards.
Frequently asked questions
How much does the Gemini 4 Neon API cost?
Google says $3 per million input tokens, $0.15 for cached input (a 95% discount) and $15 per million output tokens, with a 1M-token context window. Cached input means identical content that is sent repeatedly and billed at a lower price.
Can free users use Neon?
The channels listed in the announcement are the paid tier of the Gemini API, Gemini for business, and Google AI Pro and Ultra subscribers in the Gemini app and Antigravity; free users are not mentioned as having access.
How do I apply for Neon for cyber?
Google says you must go through the Fairwind Program and contact your Cloud account team to apply. If you are not yet a Cloud customer, the Fairwind page suggests starting with Google Cloud's onboarding documentation. Applications are vetted and approval is not guaranteed.
Are Neon's benchmark results reliable?
The results in the announcement were compiled by Google, with some drawing on third-party benchmarks. Cognition's FrontierCode 2.0 page does not list Neon, so the 46.2% figure cannot be treated as confirmed by that page.
Browse the latest news in this topic
Found this guide helpful? Share it with someone who needs it
NVIDIA and Microsoft at Ignite: Vera Rubin NVL72 in Preview on Azure, Nemotron 3.5 in Foundry, Windows 365 Direct Stream in Limited Preview
NVIDIA says it has expanded AI collaboration with Microsoft from the Azure cloud to the edge and Windows, including a Vera Rubin NVL72 preview on Azure, Nemotron 3.5 models in Foundry and a limited preview of Windows 365 Direct Stream.
Google Quantum AI Makes the Cover of Science: Quantum Contextuality Holds at Up to 105 Qubits, With a 98.3% Win Rate in a Coordination Game
Google: ARMORY CASTLE Exploited VMware vCenter Zero-Day CVE-2026-123890, Affecting vCenter 7.x and 8.x; Broadcom Has Released a Patch
Cloudflare: Election-Day Traffic to 2026 U.S. Midterm Election Sites Surged 363%, While Attackers Shifted from DDoS to Exploiting Vulnerabilities
Sources
- Introducing Gemini 4 NeonChecked:
- FrontierCode 2.0 | CognitionChecked:
- Fairwind Program — Google DeepMindChecked:
Compiled from the references and verification dates listed above; fares, schedules, product features, and terms can change, so please confirm the latest details on the relevant official website before travelling, booking, or purchasing.