Skip to content
AI NEWS#Gemini 4 Neon#Gemini 4#Gemini 4 Carbon

Gemini 4 Neon: Google's Mid-Tier Model Is Here, and Here's What It Means for You

Google didn't launch another massive model. It launched the model that most of your daily work will actually run on.

Yaakov Tzedek
Yaakov Tzedek · AI expert·6 min read
Gemini 4 Neon: the mid-tier model in Google's Gemini 4 family, November 2026
THE BOTTOM LINE
  • On November 18, 2026, Google launched Gemini 4 Neon, sitting between Carbon and Hydrogen: $3 input and $15 output per million tokens, a 1M-token context window, available to Pro and Ultra subscribers, via API, and on Antigravity.
DID YOU KNOW?

Until November 30, 2026, Gemini 4 Carbon, the heavy model, is actually cheaper via API than Neon ($2 and $10 versus $3 and $15). Only from December 1, when Carbon rises to $4 and $20, does Neon become 25% cheaper.

Google didn't launch another massive model yesterday. They launched the model that most of your daily work will probably run on.

On November 18, 2026, Google announced Gemini 4 Neon, one week after Carbon and Hydrogen. It lands exactly in the middle gap: nearly as capable as the heavy model, but faster and cheaper on every task. If you work with Gemini, whether through the app or the API, this is likely the model you'll reach for most from now on.

In short:

  • What happened: Google launched Gemini 4 Neon, the mid-tier model in the Gemini 4 family. It's available in the app for Pro and Ultra subscribers, in the paid API, and in Antigravity.
  • What it means: According to Google, performance close to leading models at a cost that is 32% lower per task compared to Carbon.
  • What to do: If you're a Pro subscriber, run one analysis task through Neon this week. If you're building on the API, run a cost comparison before December 1, when Carbon's price doubles.

What exactly did Google launch?

Neon is the third model in the Gemini 4 family to go public this month. According to Google's official announcement, "while Carbon handles the hardest and longest projects and Hydrogen handles fast everyday execution, Neon balances depth with efficiency." The stated focus: code, knowledge work (analysis, documents, spreadsheets), and scientific research.

The technical specs according to Google: a one-million-token context window, and four thinking levels (low, medium, high as default, and xhigh for the hard questions). In the Gemini app it also pulls relevant details from Gmail and Docs, like the other models.

How good is it, according to Google?

Google positions Neon as a model that reaches the level of leading models. The headline numbers it publishes: third place on the Artificial Analysis Intelligence Index (score of 61, at the xhigh level), second place on the Vals Index, and first place on the Artificial Analysis code agents benchmark (58.6). On office work benchmarks, Google claims first place on Zapier AutomationBench and on legal and tax tests.

The number that matters most to you is the cost: according to Google, at the standard thinking level, Neon costs on average 32% less per task than Carbon. It's worth keeping in mind that all these numbers come from Google itself. The real test is your task, not a table.

Where is it available, and who pays what?

In the Gemini app: For Google AI Pro and Ultra subscribers, via the model selector. According to 9to5Google, the new selector includes Auto, "everyday help" (Hydrogen), "deep analysis" (Neon), and "code and complex work" (Carbon), and Gemini 3.1 Pro is being removed from the app. Google did not detail availability by country.

In the API: On the paid tier of Google AI Studio and in Gemini for business. The official pricing: $3 per million input tokens (15 cents for cached input, a 95% discount) and $15 per million output tokens.

ModelInput (per million tokens)Output (per million tokens)
Gemini 4 Hydrogen$1.50$7.50
Gemini 4 Neon$3$15
Gemini 4 Carbon (until Nov 30)$2$10
Gemini 4 Carbon (from Dec 1)$4$20

Note the odd detail: until the end of the month, Carbon is cheaper per token than Neon, because of its launch pricing. From December 1, that flips. Full Carbon pricing details are in the article on the Gemini 4 Carbon launch.

In Antigravity: Google's development environment, for paid Pro and Ultra subscribers, in the app, the CLI, and extensions. There's an interesting new feature here: Orchestrator now lets Neon plan and Hydrogen execute within the same job. Neon builds the plan and architecture, Hydrogen makes the edits and runs the tests, and if Hydrogen gets stuck it hands the task back to Neon to fix the plan. According to Google, this saves quota. The comparison with Claude Code is in the Claude Code vs. Antigravity guide.

What does this mean for you, and what to do this week?

If you're a Google AI Pro subscriber: Open the model selector and choose "deep analysis." Take one real task from this week, an analysis of a report, a comparison of quotes, or a summary of a long document, and run it on both Neon and Hydrogen. Within ten minutes you'll know which one is worth the wait.

If you're building on the API: Before December 1, run your 20 most common requests on both Neon and Carbon, and measure real cost per task, not per token. A model that's cheap per token but thinks twice as long ends up costing more. If most of your work is repetitive, straightforward tasks, Hydrogen is probably still the right choice.

If you're developing in Antigravity: Try the Neon-as-planner and Hydrogen-as-executor pair on a medium-sized task, for example a feature with tests. This is the direction the whole industry is heading: an expensive model that thinks, a cheap model that works.

The bigger picture: Google, Anthropic, and OpenAI are no longer selling "one model." They're selling a family, and the important skill is knowing which model to send which task to. We covered the full family in the Gemini 4 Hydrogen guide and the complete Gemini guide.

Also today in leading tools

ChatGPT for Accounting: an accounting add-on with QuickBooks and Xero

On November 18, 2026, OpenAI announced on the official ChatGPT account: ChatGPT for Accounting, an Accounting plugin alongside partner integrations including Intuit QuickBooks, Xero, Oracle NetSuite, BILL, and FreshBooks. The tasks it targets: month-end reports, bank reconciliations, data entry, and audit preparation. According to reports, any change to data requires human approval before it's saved.

Availability depends on your plan and workspace settings, and in enterprise accounts the admin chooses which apps are enabled. For accountants in Israel this is less relevant right now, because local accounting software isn't on the list. What is worth doing: checking whether a process like audit preparation can flow through an export from your existing software. Source: the ChatGPT announcement on X. More on the topic in the AI for accountants guide.

Grok Bot: a new toggle that stops background checks, but it's on by default

Security firm NewCore published research on November 17, 2026, finding that two Grok Bot accounts belonging to its employees made 31 unprompted calls to Gmail, Calendar, Slack, GitHub, and Notion over five days, sometimes while the computer was off. According to NewCore, when the bot was asked directly, it replied that it had not made any calls, and xAI support confirmed this was a built-in "proactive check" feature. This is NewCore's own research, not an independent audit.

The practical takeaway: according to NewCore, on November 12 xAI added a toggle in Administrative Settings to disable Proactive Messages, but left it on by default. If you've connected work email, calendar, or code to Grok Bot, go into the settings, turn it off if you don't need it, and connect only what the bot actually requires. Source: the NewCore announcement. We covered working safely with the bot in the Grok Bot guide.

Want to work with Gemini on your real work?

A new model is only worth something if you know how to send it the right task. The Gemini video course in Hebrew includes 54 short lessons: research, emails, documents, personal assistant, and working with Google's tools, step by step, no technical background required.

To the Gemini video course

Sources: Google's announcement of Gemini 4 Neon, Neon in Google Antigravity, 9to5Google on the launch and model selector, ChatGPT on X about ChatGPT for Accounting, NewCore on Grok Bot.

Frequently asked questions

Was this article helpful?
Share:WhatsAppLinkedInX
Yaakov Tzedek
WRITTEN BY

Yaakov Tzedek

An Israeli AI lecturer, consultant and implementer who has supported 100+ organizations in AI implementation and lectured for 200+. Official content partner of Meta, Google, Canva and TikTok, featured on Israel's Channel 13.

More about Yaakov →