Introducing Gemini 4 Neon

Gemini
Google introduces Gemini 4 Neon, a frontier-level AI model optimized for fast, cost-efficient task completion across coding, enterprise, science, and cybersecurity workflows.

Summary

Google announced Gemini 4 Neon, a new frontier model designed for fast, end-to-end task completion across complex software engineering, knowledge work, and scientific workflows. Neon scores 61 on the Artificial Analysis Intelligence Index, ranks #2 on the Vals Index (70.9%) and on Arena's new Agent Arena Index, and achieves top results on benchmarks like Legal Research Bench, Tax Agent Bench, Zapier's AutomationBench, and Andon Labs' Vending-Bench 2. It excels at long-running agentic coding, ranking #1 on the Artificial Analysis Coding Agent Index (58.6) and scoring 65.2% on Terminal-Bench 4.1 and 46.2% on FrontierCode 2.0. Priced at $3 per million input tokens ($0.15 cached) and $15 per million output tokens with a 1 million-token context window, Neon offers high performance at lower cost per task, sitting on the Pareto frontier and costing roughly 32%–80% less than competing frontier models depending on thinking level. A cyber-focused variant achieves state-of-the-art results on AutoBaxBench's hard tier (76%), WebExploitBench v1.2 (51.26%), SEC-bench Pro (90.5%), and CWE-bench v2 (70.7%). On safety, Neon matches Carbon-level protections against misuse and prompt injection, records its lowest fabrication rate yet, and resists every attack in Gray Swan's IPI benchmark at its highest thinking level. Within the Gemini 4 family, Neon balances depth and efficiency between Carbon (hardest long-horizon projects) and Hydrogen (fast everyday execution). It rolls out today in the Gemini API, Google AI Studio, Gemini for business, Google Antigravity, and the Gemini app for Pro and Ultra subscribers, with the cyber variant going to trusted defenders through the Fairwind Program.

(Source:Gemini)