Back to home

Gemini 4 Neon Lands: Near-Frontier Brains at $3 per Million Tokens

Google DeepMind launched Gemini 4 Neon, a model it says ranks #3 on the Artificial Analysis Intelligence Index while costing $3 per million input tokens. For anyone building with AI or leaning on it at work, it is another sign that top-tier capability is getting cheaper by the month.

Align Times AINovember 19, 2026
CareerMoney
Gemini 4 Neon Lands: Near-Frontier Brains at $3 per Million Tokens

Source: Google Blog — View source

Google DeepMind introduced Gemini 4 Neon on November 18, pitching it as frontier-level intelligence with fast, end-to-end task completion across software engineering, knowledge work and science. According to Google's announcement, Neon scores 61 on the Artificial Analysis Intelligence Index at its highest thinking setting, good for #3 overall, just behind larger and slower models.

The price tag is the plot twist

Neon costs $3 per million input tokens and $15 per million output tokens, with cached input at $0.15 and a 1 million-token context window. That is the kind of pricing that lets a solo founder or a small team run long agent workflows without sweating the invoice.

Built for the long, boring tasks

  • Google says Neon ranks #1 on the Legal Research Bench and Tax Agent Bench, and #2 on the Vals Index at 70.9%.
  • It debuts #2 on Arena's new Agent Arena Index and #1 on Zapier's AutomationBench.
  • Google says it recorded its lowest fabrication rate yet in internal Antigravity testing.

Those are Google's reported figures and independent leaderboards, so treat them as a starting point and test on your own work.

Where you can use it

Neon is rolling out in the Gemini API in Google AI Studio, Gemini for business, Google Antigravity for paid AI Pro and Ultra subscribers, and the Gemini app for Pro and Ultra subscribers.

Why it matters: The skills being benchmarked here, spreadsheets, tax and legal research, finance modeling and multi-step automation, are exactly the tasks many early-career jobs are built from. Cheaper capable models mean the edge shifts toward people who know how to direct, check and combine them. Timing is everything: learn the tools while the gap is still an advantage. Source: Google DeepMind blog.

Related stories