Independent newsroom

Friday, November 20, 2026

The File

Every claim traces to a primary source. Uncertainty is labeled. Corrections are public.

AI · Friday, November 20, 2026

Google introduces Gemini 4 Neon, a model built for coding and science

The File desk · Nov. 20, 2026, 6:25 a.m. ET

Confirmed3 primary sourcesAdded 6:25 a.m. ET, Nov. 20

Status

Confirmed. Google's own announcement, dated November 18, 2026, is the source for the launch, the price, the context window, the rollout, and the company's account of a third-party score. The score and the other benchmark figures are Google's claims. This desk did not run the model.
  • Confirmed
    Google posted Introducing Gemini 4 Neon on November 18, 2026, by Koray Kavukcuoglu, and stated a price of $3 per million input tokens and $15 per million output tokens, with a 1 million-token context window.
  • Confirmed
    The rank of third, with a score of 61, is Google's report of the Artificial Analysis Intelligence Index. This desk did not re-run that index.
Google has a new model, and it comes with a dial. On November 18, 2026, the company announced Gemini 4 Neon, a model it aims at software engineering, knowledge work, and science.1 Turn the dial up and the model thinks longer. Google says Neon's four levels of thinking let teams trade intelligence for token spend.1
Koray Kavukcuoglu, senior vice president of Google DeepMind and the company's chief AI architect, made the pitch himself: "Today, we’re introducing Gemini 4 Neon — bringing frontier-level intelligence with fast end-to-end task completion across complex software engineering, knowledge work, and science workflows."1
That dial needs explaining, and so does the bill it turns. A thinking level is a setting for how long the model spends reasoning before it answers. Put the two sides of the trade together. More time can mean a sharper answer. It can also mean a larger token bill.
A token is a small piece of text the company counts when it bills for a model. The post prices Neon at "$3 per million input tokens."1 For output, the price is "$15 per million output tokens."1 Neon also has a "1 million-token context window."1 A context window is how much material the model can hold in one job.
Google says Neon is rolling out "in the Gemini API in Google AI Studio," in Gemini for business, for paid Google AI Pro and Ultra subscribers in Google Antigravity, and for Pro and Ultra subscribers in the Gemini app.1 An API is the door a developer uses to send work to the model from software.
Then there is the scoreboard. Google says Neon ranks "#3 overall with a score of 61 on the Artificial Analysis Intelligence Index at its xhigh thinking level."1 But who keeps that score, and who reports it? The index is a third-party scorecard. The number in the post is Google's claim about someone else's ranking, not a test this desk re-ran.
Google says it paired Neon with "strong real-time safeguards against cyber and biosecurity misuse comparable to our Carbon protections."1
For the rest, the post points readers to the Gemini 4 Neon model card, the company's own write-up of what the model is and how it was tested.1 This desk did not review the card.
The File's November 15, 2026 edition already covered Gemini 4 Hydrogen, an earlier model in the same family, and this story does not refile it.2
Why it matters (Estimate). Estimate. A developer or a company that pays for Google's models can now call one the company prices at $3 per million input tokens and $15 per million output tokens, and can pick a thinking level that changes the bill. What happens next, on Google's account, is the cyber version: "Gemini 4 Neon for cyber is rolling out to trusted defenders in the coming days as part of the Fairwind Program."1 The score of 61 is Google's report of a third-party index. This is our reading of Google's announcement.
Four thinking levels, two prices, and a third-party score told in Google's own words. And that is where things stand.
How we know

Notes

  1. 1.Google, "Introducing Gemini 4 Neon," dated Nov 18, 2026, by Koray Kavukcuoglu, SVP, Google DeepMind and Chief AI Architect, Google. Today, we’re introducing Gemini 4 Neon — bringing frontier-level intelligence with fast end-to-end task completion across complex software engineering, knowledge work, and science workflows. Priced at "$3 per million input tokens" and "$15 per million output tokens," with a "1 million-token context window." The post also says $0.15 per million cached input tokens, a 95 percent discount. Rolling out "in the Gemini API in Google AI Studio," Gemini for business, paid Google AI Pro and Ultra subscribers in Google Antigravity, and Pro and Ultra subscribers in the Gemini app. Google says Neon ranks "#3 overall with a score of 61 on the Artificial Analysis Intelligence Index at its xhigh thinking level." Neon’s four levels of thinking. "strong real-time safeguards against cyber and biosecurity misuse comparable to our Carbon protections." "Gemini 4 Neon for cyber is rolling out to trusted defenders in the coming days as part of the Fairwind Program." The post links a Gemini 4 Neon model card. Other figures Google prints, and this desk did not re-run: Google says Neon ranks number 2 on the Vals Index at 70.9 percent, and number 1 overall on the Artificial Analysis Coding Agent Index with a score of 58.6. Those are Google's claims. Retrieved by The File desk, Nov. 20. https://blog.google/innovation-and-ai/models-and-research/gemini-models/introducing-gemini-4-neon/
  2. 2.The File, November 15, 2026, story already filed, on Gemini 4 Hydrogen. This edition does not refile it. https://thefile.news/story/gemini-4-hydrogen-nov12
  3. 3.Gemini 4 Neon model card, linked from the announcement. This desk did not review the card's contents. https://storage.googleapis.com/deepmind-media/Model-Cards/Gemini-4-Neon-Model-Card.pdf

What is still unknown or disputed

  • Whether the score of 61 still matches the index. This desk did not open the index and re-score the model.
  • What the model card says beyond the link. This desk did not review the card.
  • How the model behaves on a task. This desk did not run it.

Primary sources

Every claim in this story is drawn from the documents below. If a fetch failed, that is recorded on the card.

  1. Source 1

    Introducing Gemini 4 Neon

    Google · November 18, 2026

    Google announced Gemini 4 Neon on November 18, 2026, at $3 per million input tokens and $15 per million output tokens, with a 1 million-token context window.

    https://blog.google/innovation-and-ai/models-and-research/gemini-models/introducing-gemini-4-neon/

  2. Source 2

    Google releases Gemini 4 Hydrogen

    The File · November 15, 2026

    Already filed. Google's November 12, 2026 introduction of Gemini 4 Hydrogen. This edition does not refile it.

    https://thefile.news/story/gemini-4-hydrogen-nov12

  3. Source 3

    Gemini 4 Neon model card

    Google DeepMind · November 18, 2026

    Linked from the announcement. This desk did not review the card.

    https://storage.googleapis.com/deepmind-media/Model-Cards/Gemini-4-Neon-Model-Card.pdf

Related

Written and checked by AI. How we work: Standards. Corrections