Introducing Gemini 4 Neon
Google's new mid-tier Gemini 4 Neon model is aimed at coding, enterprise analysis, and cyber defense, with a pitch of near-frontier performance at lower cost per task. The post leans heavily on third-party benchmark rankings and Google's own internal tests.
- Neon ranks #3 overall on the Artificial Analysis Intelligence Index (score 61 at its highest thinking level) and #1 on the Artificial Analysis Coding Agent Index (58.6), though Google's claims are based on its selected benchmarks.
- Pricing is $3 per million input tokens and $15 per million output tokens, with a 1 million-token context window. Google says cost per task is 32% lower than its Carbon model at the default setting.
- For cyber work, Neon scores 90.5% on SEC-bench Pro and 51.26% on WebExploitBench at about half the cost of competing large models, and Google says it cut fabrication and shortcut rates in testing.
Google's Gemini 4 Neon is pitched as a mid-tier model that punches above its weight: frontier-level results at a lower price. It ranks #3 overall on the Artificial Analysis Intelligence Index with a score of 61 at its highest "xhigh" thinking level, just behind larger, slower models, and #2 on the Vals Index at 70.9%. Google also claims #2 on steerability on the new Agent Arena Index, meaning it adjusts cleanly when a user corrects it partway through a task. Pricing is $3 per million input tokens and $15 per million output, with cached input at $0.15 (a 95% discount) and a 1 million-token context window. It's rolling out through the Gemini API, Google AI Studio, Gemini for business, Google Antigravity for paid AI Pro and Ultra subscribers, and the Gemini app for Pro and Ultra users.
The strongest claims are in enterprise and analytical work, which Google frames as spreadsheets, PDFs, charts, audio, and video handled in a single context window. Neon reportedly takes #1 on the Legal Research Bench and Tax Agent Bench, #1 on Zapier's AutomationBench (56.3%), and #1 on Andon Labs' Vending-Bench 2, a long-horizon business simulation where it earned $16,843. On coding, it ranks #1 on the Artificial Analysis Coding Agent Index at 58.6, scores 65.2% on Terminal-Bench 4.1, and reaches 39.0% on SWE-sweep inside Antigravity, against 12.8% on the mini-SWE harness. Google is clear that the Antigravity harness accounts for much of that gap, so the number reflects the scaffolding as well as the model. Scientific reasoning results include best-in-class scores on Terminal-Bench Science (70.5%), RiemannBench (84.0%), and HLE-Diamond (71.3%).
Cost and safety get as much attention as raw capability. At the default high thinking level, Google says Neon costs 32% less per task than its predecessor Carbon, and roughly 50% to 80% less than top competing frontier models on the Pareto frontier. The cyber claims are specific: state-of-the-art on the hard tier of AutoBaxBench (76%), 51.26% on WebExploitBench v1.2 at half the cost of large competitors, and 90.5% on SEC-bench Pro. Cyber capabilities are gated, rolling out to vetted defenders through the Fairwind Program. On behavior, Neon shows 0.0% cheating on Terminal-Bench 4.0 and the lowest shortcut rate of any Gemini 4 model at 2.45%, and it refuses only 1.98% of tasks on Terminal-Bench 4.0. Treat these as Google's own internal and selected-benchmark figures, not independent verification, since most of the headline rankings come from evaluations Google chose and ran or cited.