Google has released Gemini 4 Neon, a new model in its Gemini 4 family that the company says delivers frontier-level intelligence with faster end-to-end task completion. The announcement, published November 18 by Koray Kavukcuoglu, SVP of Google DeepMind and Google's Chief AI Architect, positions Neon between Gemini 4 Carbon, which handles the hardest long-horizon projects, and Gemini 4 Hydrogen, which handles fast everyday execution.
Neon is priced at $3 per million input tokens, $0.15 per million cached input tokens (a 95% discount) and $15 per million output tokens, with a 1 million token context window. It offers four thinking levels. Google says that at the default "high" level it costs 32% less per task than Carbon, and that at "xhigh" it costs at least 36% less per task than competing models at their highest settings.
On Google's reported benchmarks, Neon scores 61 on the Artificial Analysis Intelligence Index at xhigh thinking (third overall), 58.6 on the Artificial Analysis Coding Agent Index (first overall), 65.2% on Terminal-Bench 4.1, 46.2% on FrontierCode 2.0 and 1336 on the Arena Agent Arena Index (second). Google also reports 70.5% on Terminal-Bench Science, 84.0% on RiemannBench and 71.3% on HLE-Diamond. On SWE-sweep it scores 12.8% with the mini-SWE harness and 39.0% inside Google Antigravity.
The model is available through the Gemini API in Google AI Studio (paid tier) and Android Studio, in Gemini for business, in Google Antigravity for paid Google AI Pro and Ultra subscribers, and in the Gemini app model picker for Pro and Ultra subscribers. A cyber-defense variant is rolling out to vetted defenders through Google's Fairwind Program over the next few days.
On safety, Google reports a 0.0% cheating rate on Terminal-Bench 4.0, a 2.45% shortcut rate on its coding cheating evaluation (the lowest of any Gemini 4 model), a 0.09% misaligned action rate on its agentic safety suite and 0.0% attack success on the Gray Swan indirect prompt injection benchmark at its highest thinking level.
For developers, Neon is a cheaper option than Carbon for agentic coding and enterprise automation. Its cached-input price makes long-context and repeated-context workloads much cheaper. All benchmark figures are Google's own and have not yet been independently reproduced.