Google Launches Gemini 4 Neon with Frontier-Level AI Capabilities
Google has officially launched Gemini 4 Neon, a cutting-edge AI model designed to deliver frontier-level intelligence across software engineering, knowledge work, and scientific workflows. The model is now available through the Gemini API in Google AI Studio, the Gemini app, and Google Antigravity for AI Pro and Ultra subscribers.
Performance and Benchmark Rankings
Gemini 4 Neon ranks #3 overall on the Artificial Analysis Intelligence Index with a score of 61 at its xhigh thinking level, trailing only larger, slower models. It also achieves #2 on the Vals Index and lands on the Pareto frontier, demonstrating its efficiency and capability. On the Agent Arena Index, Neon ranks #2 in steerability, excelling at adapting to mid-task user corrections. Additionally, the model records its lowest fabrication rate to date in internal Google Antigravity testing.
Neon’s performance extends to specialized benchmarks, where it ranks #1 on the Legal Research Bench, Tax Agent Bench, and Zapier’s AutomationBench (56.3%). It also leads on Andon Labs' Vending-Bench 2 with a simulated business revenue of $16,843. For scientific and mathematical reasoning, Neon achieves top scores on Terminal-Bench Science (70.5%), RiemannBench (84.0%), and HLE-Diamond (71.3%).
Cost Efficiency and Accessibility
Gemini 4 Neon is priced at $3 per million input tokens and $15 per million output tokens, with a 95% discount for cached input tokens at $0.15 per million. This pricing positions Neon as a cost-effective option, offering a 25% lower cost per task compared to Anthropic’s Opus 5.5 while maintaining competitive performance. The model supports a 1 million-token context window, enabling it to handle large-scale, multi-modal inputs such as spreadsheets, PDFs, and videos.
Applications in Enterprise and Scientific Work
Neon is tailored for enterprise tasks, synthesizing data from diverse formats into structured outputs. It ranks #1 on Surge AI benchmarks for finance, accounting, and spreadsheet modeling. In Google Antigravity, Neon demonstrated its capabilities by creating a photorealistic 3D lab bench and an interactive 3D recreation of Alan Turing’s 1936 study, showcasing its ability to handle complex, multi-step workflows.
For software engineering, Neon leads the Artificial Analysis Coding Agent Index with a score of 58.6 and excels in multi-file debugging and refactoring. It also achieves state-of-the-art performance on AutoBaxBench for secure coding and WebExploitBench v1.2 for penetration testing, reinforcing its utility in cybersecurity.
Future-Ready AI for Complex Workflows
Gemini 4 Neon introduces four scalable thinking levels, allowing teams to optimize intelligence and token usage for specific workloads. At its default high setting, Neon offers a 32% lower cost per task compared to Gemini 4 Carbon, while its xhigh setting delivers deeper reasoning at a 36% cost reduction compared to competing models.
With its blend of high performance, cost efficiency, and versatility, Gemini 4 Neon is poised to redefine AI-assisted workflows across industries.