Gemini 4 Neon launches for enterprise AI tasks

A glowing neural network core embedded in a transparent server rack, surrounded by flowing data streams shaped like circuit paths, with faint outlines of global cities in the background.

Google has launched Gemini 4 Neon, a new frontier-level AI model designed for complex software engineering, scientific research, and enterprise knowledge work. It is now available via the Gemini API in Google AI Studio, Gemini for business, and for paid Google AI Pro and Ultra subscribers — but not through GitHub Copilot. The model delivers high intelligence with faster task completion and lower costs than prior-generation models, targeting tech teams in the UK, UAE, US, and Kazakhstan.

High performance at lower cost

Gemini 4 Neon ranks #3 globally on the Artificial Analysis Intelligence Index with a score of 61 at its xhigh thinking level, just behind larger, slower models. It also ranks #2 on the Vals Index and achieves top results on benchmarks for legal research, tax analysis, finance, and spreadsheet modelling. On Surge AI, it leads in accounting, finance, and constraint-following; it scores 73.5% on Chartography and 75.8% on Agents’ Last Exam v1.1. It ranks #1 on Zapier’s AutomationBench (56.3%) and Andon Labs’ Vending-Bench 2 ($16,843), a long-horizon business simulation.

Priced at $3 per million input tokens ($0.15 for cached inputs) and $15 per million output tokens, Neon costs 32% less per task than Gemini 4 Carbon at its default high setting. At xhigh, it remains 36%+ cheaper than competing frontier models at their highest settings, while delivering stronger performance on complex reasoning. With a 1 million-token context window, it handles multi-document, multi-modal inputs — including spreadsheets, PDFs, audio, and video — making it suitable for enterprise reporting, financial analysis, and technical research.

For UK and UAE enterprises focused on cost efficiency and performance, Neon offers a compelling alternative to higher-cost US models, especially in regulated sectors where auditability and structured output matter.

Secure, reliable code generation

In cybersecurity applications, Neon outperforms Carbon across the vulnerability lifecycle. On AutoBaxBench, it generates more secure code across easy, medium, and hard challenges, achieving state-of-the-art results (76%) on hard-tier tasks. It scores 51.26% on WebExploitBench v1.2 for penetration testing — double the efficiency of larger models. For code hardening, it reaches 90.5% on SEC-bench Pro (vulnerability detection) and 70.7% on CWE-bench v2 (patching), a 3.4% improvement over Carbon.

These capabilities are being rolled out to trusted defenders through Google’s Fairwind Program, starting in the coming days. This controlled release suggests Google is prioritising responsible deployment in critical infrastructure — relevant for regulated industries in Kazakhstan and the UAE.

Neon maintains Gemini 4 Carbon’s safeguards against misuse and prompt injection, while reducing false refusals: only 1.98% of tasks were declined on Terminal-Bench 4.0. It resists indirect prompt injection attacks completely on Gray Swan’s IPI benchmark (0.0% success rate). In internal evaluations, it achieved 0.0% cheating on Terminal-Bench 4.0 and a 2.45% shortcut rate — the lowest among Gemini 4 models — indicating strong alignment with user intent.

Seamless integration into enterprise workflows

Gemini 4 Neon is accessible via the Gemini API in Google AI Studio, Android Studio, Gemini for business (API and model dropdown), and Google Antigravity. Pro and Ultra subscribers can use it in the Gemini app via model selection. However, it is not integrated into GitHub Copilot, contrary to implications in some early reports.

GitHub has released updates to its Copilot app and JetBrains plugin, including support for Azure DevOps repositories, browser use in Chromium-based browsers, and GHES integration. These updates include cloud sandbox (public preview), remote access, and on-device voice dictation. But these enhancements are independent of Gemini 4 Neon and continue to rely on GitHub’s own model stack, including Copilot Fusion.

For enterprise teams, Neon’s availability across Google’s AI platforms enables tighter integration with Google Workspace, Cloud, and internal data systems. The lack of cloud sessions for GHES and Azure DevOps remains a limitation for fully autonomous workflows in air-gapped environments.

What it means
For founders and IT leaders, Gemini 4 Neon represents a step toward cost-efficient, high-integrity AI agents for production work. Its strong performance in legal, financial, and cybersecurity tasks makes it relevant for institutions in London, Abu Dhabi, and Almaty. The Fairwind Program’s phased rollout reflects Google’s caution with high-risk applications, aligning with UK and UAE regulatory expectations.

Organisations should evaluate Neon’s fit within their Google Cloud and AI Studio workflows. While it does not currently power GitHub Copilot, its API availability allows custom integrations. Teams in Kazakhstan may find value in its controlled access model and alignment with enterprise governance, supporting cautious adoption under data sovereignty considerations. As with any new AI system, testing in non-critical workflows — especially in compliance and security — is advised.

Sources

Leave a Reply

Your email address will not be published. Required fields are marked *