Google has a new AI model, and this one is built for speed as much as smarts. On Wednesday, November 18, Google introduced Gemini 4 Neon, a model it says brings “frontier-level intelligence” with fast, end-to-end task completion for software engineering, knowledge work and science.

Here is what Neon is, what Google says it can do, who can use it today and what it means for regular users.

What Gemini 4 Neon is

In its announcement on The Keyword, Google describes Neon as the middle option in the Gemini 4 family. Gemini 4 Carbon handles the hardest, longest projects, Gemini 4 Hydrogen handles fast everyday tasks, and Neon “balances depth and efficiency.”

In other words, Neon is designed for people and teams who need strong reasoning on complex work, but who also care about speed and cost.

The headline claims

Google shared a long list of benchmark results. A few of the most notable, according to the company:

  • Neon ranks No. 3 overall with a score of 61 on the Artificial Analysis Intelligence Index at its highest thinking level.
  • It ranks No. 1 on the Artificial Analysis Coding Agent Index.
  • It ranks No. 2 on the new Agent Arena Index for steerability, meaning how cleanly it adapts when a user corrects it mid-task.
  • Google says it recorded its “lowest fabrication rate yet” in internal testing.

Benchmarks are a company’s own selection of tests, so it’s worth waiting for independent hands-on reviews. Still, the focus on fewer made-up answers and better course-correction is the kind of improvement everyday users notice.

Faster and cheaper per task

Cost is a big part of the pitch. Google says Neon at its default “high” thinking level has a 32 percent lower cost per task than Carbon, and roughly 50 to 80 percent lower cost per task than top competing frontier models on certain indexes.

For developers, Google lists pricing of $3 per million input tokens and $15 per million output tokens, with heavily discounted cached input and a 1 million-token context window. That large window means it can take in long documents, spreadsheets, PDFs, audio and video in one go.

Four thinking levels

Neon offers four “thinking levels,” letting users and developers decide how much reasoning (and cost) a task needs. A quick question can run on a lighter setting, while the hardest analysis can use the top “xhigh” level for deeper, multi-step reasoning.

What it means in the Gemini app

For regular users, the most relevant changes are in the Gemini app. Google says Neon:

  • Writes with “a natural, direct tone” and well-structured responses.
  • Reads uploads like spreadsheets, PDFs and photos more precisely and turns dense material into summaries and reports.
  • Pulls in current details from the web for fast-moving topics.
  • Can run longer multi-step tasks, such as analyzing billing statements to spot why a utility charge spiked.

Safety and security

Google says Neon keeps the same protections against severe misuse and prompt injection as Carbon, while refusing fewer legitimate requests. A separate version tuned for cybersecurity defense is rolling out to vetted defenders through Google’s Fairwind Program in the coming days.

As with any AI tool, it’s wise to double-check important facts, especially for financial, legal or health questions.

Who can use it now

According to Google, Gemini 4 Neon is rolling out:

  • To Google AI Pro and Ultra subscribers in the Gemini app, through the model drop-down.
  • To developers through the paid Gemini API in Google AI Studio and Android Studio.
  • To businesses through Gemini for business and Google Antigravity, Google’s agent development platform.

The wrap-up

Gemini 4 Neon shows where the AI race is heading: not just bigger models, but models that do more per dollar and per second. If Google’s numbers hold up in real-world use, Neon could become the default choice for people who want strong reasoning without waiting or paying for the largest model.

Image: Laptop Code by Matthew Henry via StockSnap, CC0 1.0