The full story · 2 min

Google now sells four Gemini 4 models, from the fast Hydrogen to the strongest, Carbon. On November 18 it added Neon in the middle, and an outside test group scored it on the same day.

What Google released

Google released Gemini 4 Neon on November 18. It is in the Gemini API for paying developers and in Gemini for business. Paying Google AI Pro and Ultra subscribers also get it in the Gemini app and in Antigravity, Google’s coding app.

Neon costs $3 per million input tokens and $15 per million output tokens, with a 95% discount for cached input. It can read up to 1 million tokens at once. Google did not say which countries get it.

9to5Google reports that Neon replaces Gemini 4 Argon in the Gemini app’s model menu. Google’s own post does not say so.

Sources134

Record
Where Gemini 4 Neon fits, and what it costs.

From Google’s post of November 18 and Artificial Analysis’s tests of the same day.

The outside test

Artificial Analysis, an independent test group, scores Neon 61 on its Intelligence Index at its highest effort setting. That puts it third, behind GPT-6.2 Astra and Claude Fable 5.2. Gemini 4 Carbon scores 60 at its high setting.

On the group’s Coding Agent Index, Neon scores 59 in Google’s Antigravity tool. Claude Opus 5.5 at its highest setting scores 58. A task cost $6.86 on average with Neon and $12.61 with Opus 5.5, about half as much.

Sources2

The price catch

Carbon costs $2 per million input tokens and $10 per million output tokens until November 30, then $4 and $20. So Neon costs 50% more per token than Carbon today, and 25% less from December 1.

Running the group’s whole Intelligence Index cost $6.44 with Neon and $7.76 with Carbon at its full price, about 17% less, Artificial Analysis says. Neon was also faster, at about 21 minutes against about 36.

Sources24

What is still Google’s claim

Google lists more than a dozen of its own test results in coding, science and security, and they have not been checked outside. Its own table shows Neon slightly behind Opus 5.5 on two coding tests, Terminal-Bench 4.1 and FrontierCode 2.0, OfficeChai reports.

Google also says Neon ranks second on the Vals Index, a test of business tasks run by another company, with 70.9%. Vals has not yet published its own page for Neon.

Sources15

Sources

Every fact in this story comes from the sources below. Open them to check our work.

  1. 1
    Primary source · November 18, 2026Introducing Gemini 4 Neon Google
  2. 2
    Research · November 18, 2026Google continues to accelerate its release cadence with Gemini 4 Neon Artificial Analysis
  3. 3
    Research · November 18, 2026Google rolls out Gemini 4 Neon 9to5Google
  4. 4
  5. 5
    Research · November 18, 2026Gemini 4 Neon benchmarks OfficeChai

AI desk: Security and frontier models · published November 20, 2026, 14:44 CET. This story was generated by AI systems under the editorial rules of Silicon AI News and checked against the sources it links. No person read it before it was published. The site’s owner and editor: Y. D. Daniel, Budapest, Hungary. Found a mistake? Email editor@siliconainews.com. How we report

We count only the page and your answer.Thank you. We counted only this page and your answer.
More: words to know, tags and how to cite this story
Words to know
Token
A small piece of text, about three quarters of a word, that AI companies use to count and charge for work.
Intelligence Index
Artificial Analysis’s combined score from many tests, used to rank AI models against each other.
Cost per task
What it costs in total to finish one test job, which depends on price and on how many tokens a model uses.
Basics explainedWhat is an AI model? What happens when you use an AI model?

Google’s Gemini 4 Neon ranks third on an outside test. Silicon AI News, November 20, 2026. https://siliconainews.com/stories/google-gemini-4-neon-launch