Jiberish

AI News

Gemini 4 Neon: when it's worth twice the price of Hydrogen

Google shipped Gemini 4 Hydrogen and Neon a week apart. Where Neon fits between Hydrogen and Carbon, what it costs on each plan, how to read the launch benchmarks, and a one-week test to run.

Kyle Del FranciaUpdated 7 min read
In this piece

Google released Gemini 4 Neon on November 18, six days after Gemini 4 Hydrogen and eight days after Carbon reached general availability. That makes four Gemini 4 models in seven weeks, and the Gemini app's model picker has changed with almost every one. If your team uses Gemini, the useful question isn't which model scored highest. It's which one to pick by default, and when to move up. This article covers the Gemini app, Gemini for business, Google Antigravity, and the Gemini API. We checked Google's sources for these releases on November 20, 2026. Google is changing this lineup week to week, so confirm what your own account shows.

What changed this month

The Gemini 4 family now has three working tiers, and Google describes each one plainly (Neon launch):

ModelGoogle's descriptionAPI price per 1M tokens (input / output)
HydrogenFast, everyday execution$1.50 / $7.50
NeonComplex coding, analytical reasoning, and production business workflows$3.00 / $15.00
CarbonThe hardest, long-horizon projects$2.00 / $10.00 through November 30, then $4.00 / $20.00

Hydrogen replaces Gemini 3.8 Flash. It's available on every AI Studio tier and in the Gemini app, including for free users through Auto, and it now powers Google Search for everyone (Hydrogen launch).

Neon is adapted from Carbon (model card) and takes over from Argon, the first Gemini 4 model. Google's developer guide names Neon as the upgrade from both Argon and Gemini 3.1 Pro (developer guide). In the Gemini app, 9to5Google reports the picker now reads Auto, Everyday help (Hydrogen), Deeper analysis (Neon), and Code & complex work (Carbon), with Argon and 3.1 Pro gone (9to5Google). Neon has a 1M-token context window and four thinking levels, with high as the default.

Only ten days ago, Google said Argon "remains the ideal choice" for most software and knowledge work. If someone on your team wrote that down as a rule, it's out of date.

Where Neon fits

Our default for a team on Gemini: Hydrogen for quick, high-volume work, Neon for anything that needs real reasoning, and Carbon only for the tasks where Neon falls short. That ladder follows Google's own positioning and prices. It's our starting rule, not a result from our own testing yet.

Neon earns its place on work with several steps and a real cost of getting it wrong: a pricing analysis across three spreadsheets, a contract comparison, a code change that touches many files, or a multi-step task in the Gemini app. Google says Tasks in the Gemini app can now run on Neon, and that it reads uploaded spreadsheets, PDFs, and photos more precisely (Neon launch). For drafting a reply, summarizing a thread, or answering a routine customer question, Hydrogen will do the job at half the per-token price.

Google also says Neon costs 32% less per task than Carbon at the default thinking level. That's the right way to think about cost, comparing what a finished task costs rather than the price per token, but it's Google's measurement on Google's task mix. One wrinkle matters this week: until November 30, Carbon's introductory price is lower per token than Neon's. From December 1, Neon is 25% cheaper per token than Carbon and Argon.

How to read the launch benchmarks

Google's launch post is full of rankings. It puts Neon third on the Artificial Analysis Intelligence Index, "just behind larger, slower models," second on the Vals Index at 70.9%, and first on the Artificial Analysis Coding Agent Index (Neon launch). Some of those are third-party leaderboards, but Google chose which ones to show, and the post reports them itself.

Treat them as a reason to test Neon, not a reason to switch to it. A benchmark tells you how a model did on someone else's tasks. Your proposals, spreadsheets, and customer emails aren't on any leaderboard.

What it costs on your plan

Where you use GeminiNeon access
Free Gemini app and Google AI PlusNo Neon. Hydrogen
Google AI Pro and UltraNeon in the Gemini app model picker, in Antigravity, and in AI Studio
Gemini for businessOff by default in the app until an admin turns it on; on by default in AI developer tools
Gemini APIPaid tier only, $3 / $15 per 1M tokens, $0.15 for cached input

The rows draw on Google's Neon launch post, Gemini subscriptions, Gemini Enterprise release notes, and AI Studio's Neon migration page; plan prices are on The Bench. If you run Gemini for business, nobody on your team sees Neon in the app until you turn it on, and the same goes for Hydrogen. Google's post makes no Workspace-specific claims for Neon, so if your team uses Gemini through a Workspace business account, check what your own model menu shows before planning around it.

Check the data terms first

The model doesn't change the data terms. The account does. Google says chats in the consumer Gemini app can be used to improve its models and read by human reviewers when activity is kept, while Workspace and paid API use have business terms. Our Google page on The Bench lays out the details. If people on your team are trying Neon on a personal AI Pro plan, keep client data out of it.

A one-week test

Here's the test we'd run this week. It's our recommendation, not something Google prescribes.

  1. Pick two tasks your team found hard in the last month, the kind that took several rounds of correction. Strip out client names, account numbers, and anything confidential.
  2. Run each task on Hydrogen and on Neon, with the same prompt and the same attached material. If you have Carbon, run it there too; this is the cheapest week to compare on the API.
  3. Note the thinking level each run used. On the API, Neon defaults to high.
  4. For each run, record how long it took, how many corrections the output needed, and whether a reviewer could trace every claim back to the material.
  5. Decide per task, not per person. If Neon saved real correction time on one kind of work, write it down as the rule ("Neon for proposal analysis, Hydrogen for everything else") and share it.

If Hydrogen held up on both tasks, that's a useful answer too. Stay on the cheaper model.

For developers on the API

The model ID is gemini-4-neon. Coming from Carbon, Argon, or Hydrogen, AI Studio's migration page says only the model ID changes, with one catch: Neon's default thinking level is high, where Carbon and Argon default to medium. Set medium explicitly if you need the old latency. Coming from 3.8 Flash or older takes more work (developer guide):

  • thinking_budget and MINIMAL return an error; use thinking_level.
  • Remove temperature, top_p, and top_k, and drop explicit context caching, which Neon doesn't support.
  • On the Agent Platform, Neon runs in global, us, and eu. Argon users in europe-west2 or asia-southeast1 have to move.

Google also designed Neon and Hydrogen to work together. On the Interactions API, the guide has Neon plan or check the work and hand individual steps to Hydrogen, and Antigravity's Orchestrator now pairs them the same way in one session (Antigravity). For an agent that runs many small steps, that split is worth testing before you put every call on Neon.

Where this leaves Gemini

Gemini stays on Watch on The Bench, where the full lineup, prices, and data terms are. Neon makes the case stronger, but the business pieces we're waiting on haven't landed: skills reach Scheduled Release Workspace domains on November 30, and when we last checked, the Gemini agent had no published plans or price. The short version for teams already on Gemini: make Neon the default for hard work, keep Hydrogen for the rest, and move up to Carbon only when a specific task needs it.

If you'd like help setting model rules for a team, see AI training for teams.

Put this into practice

Bring the decision you’re working through.

A free discovery session is a chance to review your goal, the tools or website you have, and the help you need. We’ll identify a sensible next step and discuss scope before any project commitment.

AI Advisory & Training →Custom AI Systems & Agents →AI Integration →Book a free call