Skip to main content
The Quantum Dispatch
Back to Home
Cover illustration for Gemini 4 Neon: Carbon-Class Smarts at $3 Per Million Tokens

Gemini 4 Neon: Carbon-Class Smarts at $3 Per Million Tokens

Gemini 4 Neon scores 61 on the Artificial Analysis index at $3 per million input tokens and cuts cost per task 32% versus Carbon. See who should switch.

Dr. Nova Chen
Dr. Nova Chen★Nov 20, 2026★4 min read

Google has added a fourth member to its Gemini 4 family, and Gemini 4 Neon may be the most practical one yet. Released on November 18, Neon is pitched as frontier-level reasoning at a lower price and higher speed, sitting between the fast Hydrogen model and the heavyweight Carbon. The interesting part is not that a new model arrived. It is that Google's own chart now shows a mid-tier model edging past its flagship on a major independent index while costing less per task.

  • Price: $3 per million input tokens, $0.15 for cached input and $15 per million output tokens, with a 1 million-token context window.
  • Independent score: 61 on the Artificial Analysis Intelligence Index at its highest thinking level, versus 60 for Carbon, 54 for Hydrogen and 50 for Argon, according to The Register.
  • Cost per task: Google says Neon at its default setting costs 32% less per task than Carbon.
  • Availability: 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 users.

What Is Gemini 4 Neon?

Gemini 4 Neon is a reasoning model built for complex coding, analysis and long-running agent work that has to happen quickly and often. Google DeepMind's Koray Kavukcuoglu describes the split this way: Carbon handles the hardest long-horizon projects, Hydrogen handles fast everyday execution, and Neon balances the two. Like Hydrogen, it offers four thinking levels, from low to extra high, so teams can trade depth for token spend per workload.

In the Gemini app, 9to5Google reports that Neon replaces Argon and retires the older 3.1 Pro option. The model picker now offers Auto, Hydrogen for everyday help, Neon for deeper analysis, and Carbon for code and complex work.

How Do Gemini 4 Neon Benchmarks Compare?

The independent numbers are the place to start. Artificial Analysis ranks Neon third overall on its Intelligence Index with a score of 61, just behind larger and slower models, and first on its Coding Agent Index at 58.6. Google also highlights second place on the Vals Index at 70.9% and a 1336 score on Arena's new Agent Arena Index, built from more than 4,200 real agent sessions.

The rest of the long benchmark list comes from Google and should be read as the vendor's claims. Those include 65.2% on Terminal-Bench 4.1, first place on Zapier's AutomationBench at 56.3%, and top marks on legal research and tax agent tests. Google also reports its lowest fabrication rate yet in internal Antigravity testing, which matters more to working teams than most leaderboard wins.

Gemini 4 Neon Pricing and Where It Fits

At $3 input and $15 output, Neon costs twice as much as Hydrogen. The comparison with the top tier is more favorable. Carbon and Argon run at a $2 input promotional rate until November 30, then move to $4 input and $20 output, which leaves Neon about 25% cheaper on list price, as 9to5Google notes. Factor in Google's 32% lower cost per task, and Neon looks like the default choice for many production workflows. Our breakdown of Gemini 4 Hydrogen's price and speed covers the budget end of the range.

Neon for Security Teams and Safety Results

Google is also aiming Neon at defenders. The company reports 76% on the hard tier of the AutoBaxBench secure-code test and 90.5% on SEC-bench Pro for finding vulnerabilities, and says a cyber-tuned version reaches trusted defenders through its Fairwind Program in the coming days. On the safety side, Google says Neon keeps Carbon-level protections while refusing only 1.98% of legitimate Terminal-Bench 4.0 tasks, and resisted every attack in Gray Swan's indirect prompt injection benchmark with real-time safeguards switched on.

Should You Switch to Gemini 4 Neon?

If you already route between models, Neon is worth a test anywhere Argon or Carbon currently handles routine analysis. The Register points out that the element names do not map neatly onto a hierarchy, and Antigravity's Orchestrator can swap models automatically, so pinning a model for predictable billing is sensible while you measure. Teams that depend on the newest Carbon features for very long projects can keep it there. For more model releases, follow our AI coverage.

Sources: Google — November 18, 2026; The Register — November 19, 2026; 9to5Google — November 18, 2026.

More AI Stories