Google released Gemini 4 Neon, a new artificial intelligence model aimed at software engineering, office knowledge work and scientific tasks, on 18 November 2026. Koray Kavukcuoglu, Senior Vice President at Google DeepMind and Google's Chief AI Architect, announced the model in a post on Google's official blog. Google pitches Neon as a faster, cheaper model that still performs close to the most capable systems.

Who can use Gemini 4 Neon, and what does it cost?

Developers can use Neon from 18 November through the Gemini API in Google AI Studio and through Gemini for business. Paying Google AI Pro and Ultra subscribers can use it in Google Antigravity, Google's coding tool, and in the Gemini app. Google prices the model at $3 per million input tokens, $0.15 per million cached input tokens and $15 per million output tokens. It can take in up to 1 million tokens of context, enough to read large spreadsheets, PDFs, charts, audio and video together.

What Google claims about performance

All the benchmark figures below come from Google's announcement and have not been independently checked by MapMyTimes. Google says Neon scored 61 on the Artificial Analysis Intelligence Index at its highest "xhigh" thinking setting, which puts it third overall, behind larger and slower models. It also says Neon:

  • ranks first on the Artificial Analysis Coding Agent Index with 58.6, and scores 65.2% on Terminal-Bench 4.1;
  • ranks second on the Vals Index at 70.9%, and first on the Legal Research Bench and Tax Agent Bench;
  • scores 84.0% on RiemannBench and 71.3% on HLE-Diamond for mathematical and scientific reasoning;
  • debuts second, with 1,336 points, on Arena's new Agent Arena Index, which draws on more than 4,200 real-world agent sessions.

Neon offers four "thinking levels". Google says that at the default "high" level its cost per task is 32% lower than Gemini 4 Carbon's, and roughly 50% to 80% lower than top rival frontier models on two independent indexes.

Is there a cybersecurity version?

Yes. Google says a "Gemini 4 Neon for cyber" variant writes more secure code than Carbon and scored 90.5% on SEC-bench Pro, a test of finding software vulnerabilities. That version is not public. Google says it will reach "trusted defenders" in the coming days through its Fairwind Program.

What about safety?

Google says Neon keeps the same protections against severe misuse and prompt injection as Gemini 4 Carbon, with real-time safeguards against cyber and biosecurity misuse. According to the company, the model took shortcuts on 2.45% of tasks in its internal coding-cheating test, the lowest of any Gemini 4 model, and passed 99.2% of its agent honesty evaluation. Google has also published a model card with more detail.

Why it matters

The launch continues a busy week of AI releases from Google. MapMyTimes earlier reported that the company made Gemini's Tasks mode free in six countries, including India. Neon's pricing and claimed results put it in direct competition with rival models released in recent weeks. Users should expect independent tests to confirm or challenge the company's numbers.

Photo source · Negative Space, CC0