Google Launches Gemini 4 Neon for Enhanced Coding and Research Efficiency

Google has launched Gemini 4 Neon, a new model aimed at improving performance in complex software engineering, scientific research, and enterprise workflows. The company claims it can achieve task completion three times faster than its larger Gemini 4 Carbon model while maintaining comparable quality.
Gemini 4 Neon is tailored for long-running, multi-step tasks such as debugging large codebases, analyzing extensive datasets, and coordinating multiple agents to complete intricate projects. According to Google, Neon matched or exceeded Carbon in overall quality during early internal evaluations while completing tasks three times faster.
The model scored 61 on the Artificial Analysis Intelligence Index, placing it third overall, and ranked second on the Vals Index and the new Agent Arena Index, which measures how effectively a model adapts to user corrections during tasks.
Performance Metrics
Software engineering is a primary focus of Gemini 4 Neon. Google reported that the model ranked first on the Artificial Analysis Coding Agent Index with a score of 58.6 and achieved 65.2 percent on Terminal-Bench 4.1. It also scored 39 percent on SWE-sweep when tested within Google’s Antigravity environment, significantly improving from its previous 12.8 percent result on the mini-SWE harness.
These capabilities are designed to assist developers in managing multiple files, debugging complex repositories, and making targeted changes without the need to repeatedly process the same information. Neon can also function as an orchestrator in Google Antigravity, directing parallel Hydrogen subagents to handle different aspects of a task.
Applications Beyond Coding
Beyond coding, Google is positioning Neon for enterprise analysis and scientific research. Its 1-million-token context window allows it to process large volumes of information across various formats, including spreadsheets, PDFs, charts, audio, and video. The model has ranked first in several evaluations, including the Legal Research Bench, Tax Agent Bench, and Zapier’s AutomationBench.
Neon also achieved 70.5 percent on Terminal-Bench Science and 84 percent on RiemannBench, which tests mathematical reasoning. These features are expected to enhance its utility in diverse fields.
Cost Efficiency and Cybersecurity Features
Google is emphasizing Neon’s cost-effectiveness and cybersecurity capabilities. The model is priced at $3 per million input tokens and $15 per million output tokens, with cached input tokens costing $0.15 per million, representing a 95 percent discount. At its default high thinking level, Neon reportedly offers a 32 percent lower cost per task compared to Carbon.
For cybersecurity, a specialized version of Neon is designed to help defenders identify vulnerabilities, validate risks, and patch weaknesses in software. It scored 90.5 percent on SEC-bench Pro and 76 percent on the hard tier of AutoBaxBench, according to Google.
Access to this cybersecurity version will initially be limited to trusted defenders through Google’s Fairwind Program. Gemini 4 Neon is being rolled out through the Gemini API in Google AI Studio and Gemini for business, with availability for Google AI Pro and Ultra subscribers through the Gemini app and Antigravity.
The launch of Gemini 4 Neon adds another model to Google’s Gemini 4 family, positioned between the faster Hydrogen model and the more powerful Carbon model, catering to demanding, high-frequency workloads.
Source: Interesting Engineering