
Gemini 4 Neon Is on the $19.99 Plan: What Ultra's Extra $80 Buys
- OrcaNEWOrcaCyber: Hy4 Flash Abliterated / Uncensored2026-11-18$0.07 / $0.28 per 1M tokens · 184 tok/s
- tencentNEWTencent: Hy4 Flash2026-11-16$0.02 / $0.07 per 1M tokens · 494 tok/s
- AlibabaNEWQwen: Qwen4 Max2026-11-16$1.50 / $4.50 per 1M tokens · 385 tok/s
- GoogleNEWGoogle: Gemini 4 Carbon2026-11-10$2.00 / $10.00 per 1M tokens · 94 tok/s
- AnthropicNEWAnthropic: Claude Fable 5.22026-11-09$8.00 / $40.00 per 1M tokens · 290 tok/s
- OrcaOrcaCyber: GLM 5.3 Abliterated / Uncensored2026-11-04$3.00 / $5.00 per 1M tokens · 99 tok/s
- OrcaOrcaCyber: Hy4 Abliterated / Uncensored2026-11-04$0.80 / $1.60 per 1M tokens · 132 tok/s
- OrcaOrcaCyber: Kimi K3.1 Pro Abliterated / Uncensored2026-11-04$2.00 / $5.00 per 1M tokens · 86 tok/s
- tencentTencent: Hy42026-11-04$0.83 / $2.50 per 1M tokens · 122 tok/s
- OrcaOrcaCyber: MiniMax M3.1 Flash Abliterated / Uncensored2026-10-29$0.80 / $1.60 per 1M tokens · 104 tok/s
- OrcaOrcaCyber: GLM 5.3 Flash Abliterated / Uncensored2026-10-27$0.80 / $1.60 per 1M tokens · 66 tok/s
- OrcaOrcaCyber: Qwen3.8 Flash Next Abliterated / Uncensored2026-10-26$0.80 / $1.60 per 1M tokens · 89 tok/s
- OrcaOrcaCyber: MiMo V2.6 Flash Abliterated / Uncensored2026-10-26$0.80 / $1.60 per 1M tokens · 41 tok/s
- MoonshotMoonshotAI: Kimi K3.1 Pro2026-10-2447Intelligence
- OrcaOrcaCyber: DeepSeek V4.1 Flash Abliterated / Uncensored2026-10-23$0.80 / $1.60 per 1M tokens · 215 tok/s
- minimaxMiniMax: M3.1 Flash2026-10-1929Intelligence
- OrcaOrca: OrcaCyber Ultra 1.02026-10-17$5.00 / $10.00 per 1M tokens · 109 tok/s
- deepseekDeepSeek: DeepSeek V4.1 Flash 10172026-10-1733Intelligence
- DeepSeekDeepSeek: DeepSeek V4.1 Flash 09102026-10-17$0.15 / $0.60 per 1M tokens · 160 tok/s
- DeepSeekDeepSeek: DeepSeek V4.1 Flash2026-10-1740Intelligence
On the evening of 20 November, Chubby — the account @kimmonismus — posted a screenshot with one sentence attached: "Yes, Gemini 4 Neon is that good. I subscribed to Ultra tier." By the time we read it the post had 1,216 likes. It is the kind of testimonial a model launch wants, and it leaves open the question the reply thread immediately asked: do you have to pay $99.99 for Gemini 4 Neon? The post doesn't say so, and Google's own pages say no. Neon is not an Ultra model. It has been included on the $19.99 Google AI Pro plan since it went generally available on 18 November 2026, alongside Google AI Ultra and Gemini Enterprise, and it is the default model for new conversations in Google Antigravity on every plan that has it. Buying Ultra does not get you Neon. Buying Ultra gets you more of Neon — plus one thing Pro genuinely cannot reach.
That distinction is the whole purchase decision, and it is buried in marketing copy that says "access to Gemini 4 Neon" in one column and "higher access to Gemini 4 Neon" in the next. What follows is the arithmetic of the upgrade, using only figures Google publishes and the two products' own documentation, read on 21 November 2026.
Neon is a $19.99 model, not a $99.99 one
Google One's plan comparison, as it stands today, has a model-access row for each of the three Gemini 4 tiers it sells inside its consumer products. Google AI Plus — $9.99 a month — gets exactly one of them: "access to our Hydrogen model." Google AI Pro, $19.99, adds a second: "access to our Gemini 4 Neon model." Google AI Ultra, $99.99, checks all three, including Gemini 4 Carbon, and adds a word in front of the other two — "higher access to Gemini 4 Neon for deeper analysis in the Gemini app," and "higher access to Gemini 4 Neon in Google Antigravity and Google AI Studio."
The word doing the work in that sentence is higher, and it is a quantity, not a permission. Pro already has Neon. Ultra has a larger allowance of it.
Antigravity's documentation is blunter, because it has to be — it publishes a table. Gemini 4 Neon is marked unavailable for Free and Google AI Plus, available for Google AI Pro, available for Google AI Ultra, available for Enterprise. The same page notes Neon sits "at the top of the model selector on every plan that includes it" and that it is "the default model for new conversations if you haven't chosen a model." A Pro subscriber who never opens a settings menu is already using Neon.
Where the plans genuinely diverge on models is Carbon. Google AI Pro does not get Gemini 4 Carbon in the app or in Antigravity; Ultra does. If you want the strongest reasoning model in Google's consumer stack — the one Google's Neon announcement benchmarks Neon against in nearly every paragraph — Ultra is the only in-product route to it. Neon is not.
• Neon in the Gemini app — Pro yes, Ultra yes, Plus no, Free no.
• Neon in Antigravity — Pro yes, Ultra yes, Enterprise yes, Plus and Free no.
• Carbon in the Gemini app — Ultra only.
• Carbon in Antigravity — Ultra only.
• Hydrogen — Plus, Pro and Ultra; the only Gemini 4 tier a $9.99 plan touches.
• Orchestrator in Antigravity — Ultra and paid Pro.


What the extra $80 actually buys
The gap between Google AI Pro and the 5x Google AI Ultra is exactly $80 a month: $99.99 against $19.99, and Ultra is billed monthly only, with no annual option on the page. Google does not publish a price for the Ultra 20x tier anywhere we could find it — the comparison table has the column, the price cell is empty, and the plan card reads "From /mo" — so every number below is about the 5x tier.
What that $80 is not buying is access. What it is buying, per Google's own support documentation, is quota: the Gemini apps publish "compute-based usage limits" that "factor in the complexity of your prompt, the models and features you use, and the length of your chat," refresh every five hours, and stop at a weekly cap. The tiers are relative, not absolute — Free is "standard limits," Plus is twice that, Pro is four times it, and Ultra is "5x or 20x higher than AI Pro limits depending on your subscription." Antigravity describes the same shape from the other end: Ultra gets "the highest, most generous quota, refreshed every five hours" and "the highest weekly rate limits," Pro gets "a high, generous quota, refreshed every five hours until the weekly limit is reached," and users on neither plan get "a meaningful quota, refreshed weekly."
Because the limit is compute rather than a message count, there is no honest way to convert $80 into "more Neon." But there is an honest equivalence to hold next to it, using the API's published rate card. At $15.00 per million output tokens and $3.00 per million input tokens, $80 is what 5.33 million output tokens or 26.7 million input tokens cost on gemini-4-neon. At a 3:1 input-to-output blend — the ratio Artificial Analysis uses for its blended-price column, where Neon lands at $6.00 per million — $80 is about 13.3 million tokens. That is the API value of the money. The subscription spends it on a compute allowance Google deliberately declines to denominate in tokens, which is either generous or opaque depending on how your month goes.
• Google AI Pro — $19.99/mo; Neon and Hydrogen in-app; 4x the free compute allowance; 1M-token context in the app.
• Google AI Ultra (5x) — $99.99/mo; Neon, Carbon and Hydrogen in-app and in Antigravity; 5x the Pro allowance; 1M-token context in the app.
• Google AI Ultra (20x) — price not published on Google's plan page; 20x the Pro allowance, 30 TB storage, 2,000 Colab units a month, and the only tier that includes Project Genie.
• Overages — Pro and Ultra can buy AI credits past the cap, billed "at standard Gemini Enterprise consumption pricing," controlled by an AI Credit Overages setting with two values, Never and Always.
If your usage already fits Pro's four-times-standard allowance, the $80 buys nothing you will touch, and the tweet's author bought a bigger bucket rather than a better model. If you are routinely hitting the five-hour refresh and the weekly cap, the same $80 buys five times the headroom for the model you were already using — which is a defensible purchase, just not the one the post describes.
Three products, two context windows
The other thing the plans change is how much of Neon you can put in front of it at once, and here Google's own pages disagree in a way that matters for anyone planning a long-context job.
The API changelog announces gemini-4-neon with "a unified 1M-token context window." The Gemini apps support page lists 1 million tokens for both AI Pro and AI Ultra, 128k for AI Plus and 32k for free. Artificial Analysis's page for the model, which we read on 19 November, also states 1M. Antigravity's model documentation says something different: "Context window: Up to 512K tokens."
So the same model, reached three ways, gives you two ceilings. This is not a plan difference — Antigravity's 512K applies on Pro and Ultra alike — it is a surface difference, and it is the sort of detail that turns into an afternoon of debugging when an agent harness silently truncates a repository you thought fit.
• Through the Gemini API — 1M tokens, at $3.00/$15.00 per million, with batch and flex at half rate and priority at 1.8x.
• In the Gemini app — 1M tokens on Pro and Ultra, 128k on Plus, 32k free.
• In Antigravity — up to 512K tokens, on every plan that has Neon.
• Thinking levels — low, medium, high (default) and xhigh, spelled Extra in Antigravity's selector.
• Caching — implicit only; explicit caching is not supported on any Gemini 4 model, and cached input bills at $0.15 per million.
The 512K figure deserves one more sentence because it is the least-reported number on the page and the one that most changes an Antigravity workflow: a Pro subscriber and an Ultra subscriber running the same agent get the same context ceiling. What they do not get is the same number of attempts before the quota resets.

The limits are compute, not message counts
This is the part most upgrade decisions get wrong, because the two products meter differently and neither publishes a number you can plan against.
In the Gemini app, the meter is compute-based and resets on a five-hour cycle inside a weekly cap. That means two users on the same plan can exhaust it at wildly different rates: a long chat with a large file attached costs more than a dozen short ones, and Google says the complexity of the prompt is one of the factors. A plan comparison built on "prompts per day" would be fiction, which is presumably why Google doesn't publish one.
In Antigravity, the same five-hour-and-weekly shape applies, but the quota is split by model family — the product shows separate "Gemini Models" and "Claude and GPT models" allowances, each with a weekly and a five-hour remaining figure. Neon draws on the Gemini side. A subscriber who spends their week in Codex or Claude Code inside Antigravity is not spending the Neon allowance at all.
Which produces the only clean rule in this piece: the upgrade is worth it if, and only if, you can point at a specific week in which you hit the Pro cap on Gemini models and had work left. Everything else is buying a bucket for rain you have not seen.
What an Ultra subscriber's verdict is worth
Almost nothing as evidence about the model, and something as evidence about the plan.
The post is one screenshot, no benchmark, no comparison and no method — a subscriber saying he is impressed and that he upgraded. We have no way to know whether the $80 went to Neon usage, to Carbon, to the video and image generation models Ultra adds, to Project Genie on the 20x tier, or to the 20 TB of storage. The account is a well-followed one and the enthusiasm reads as genuine; it is still a single data point with no denominator.
What it does demonstrate is that the marketing has landed. A technically literate subscriber came away believing Neon is an Ultra feature, and then paid for Ultra. That is the outcome Google's copy is built to produce — "higher access" in the Ultra column, "access" in the Pro column, same model name in both — and it is why the plan question keeps coming up in replies to the launch posts.
If you want a number rather than a testimonial, the independent one is on Artificial Analysis: Gemini 4 Neon at its xhigh setting scores 61.4 on Intelligence Index v5.0.1, sixth of 96 models, against 59.6 at the high setting it actually defaults to, and 60.0 for Gemini 4 Carbon at Carbon's own maximum. Pushed to a rung Carbon does not have, Neon wins by 1.4; compared at the settings each vendor recommends, it loses by 0.4. Those are the figures to put next to a $19.99 subscription, not a like count.
If you want the API instead of the app
There is a third option the tweet doesn't consider, and for anyone doing programmatic work it is usually the right one: skip both plans and pay per token.
The API carries no subscription and no five-hour reset. Neon at $3.00/$15.00 with batch at $1.50/$7.50 means a workload that runs in batch pays exactly the per-token rate with no allowance to exhaust. The catch is that the rate card is now the thing to plan around: Gemini 4 Carbon and Gemini 4 Argon both sit at $2.00/$10.00 through 30 November 2026 and step to $4.00/$20.00 on 1 December, while Neon's card does not move. A December budget built on November's Carbon price is wrong by a factor of two, and that includes budgets built on the Ultra subscription's in-app Carbon access, which is a different meter entirely.
This is also where the routing question becomes concrete rather than promotional. Gemini 4 Carbon and Gemini 4 Argon are on OrcaRouter's catalogue at provider list price with no markup, which means the 1 December step to $4/$20 is live here the same day it is live at Google — no repricing lag, no second contract. Gemini 4 Neon is not in our catalogue, so nothing in this section is an argument for calling Neon through us; it is an argument for not building a production path on a single vendor's plan terms while you decide.
For the evaluation itself, the plan-vs-API choice is reversible and the harness choice is not. A model you have not measured in your own repository is a model you should be able to swap out of a request without a code change. That is what a routing layer is for: one endpoint, a fallback chain that can put Argon behind Carbon behind whatever you are testing, and automatic failover when a provider's five-hour window closes mid-run.
Who should upgrade, and who shouldn't
Upgrade to Google AI Ultra if what you want is Carbon inside the Gemini app or Antigravity — that is the one model the $80 genuinely unlocks in-product — or if you have measured yourself hitting Pro's five-hour Gemini cap and need five times the ceiling. The 20x tier adds Project Genie, which nothing else sells, at a price Google does not publish on the comparison page; if that is your reason, get the number from the checkout flow before you plan around it.
Stay on Google AI Pro if you want Gemini 4 Neon and nothing else. You have it. It is the default model in Antigravity and it is in the app's model dropdown, at the same 1M-token app context the Ultra tier gets, on the same 512K Antigravity ceiling, with the same four thinking levels. The only thing you are short of is quantity.
Skip both and use the API if your work is programmatic, batchable, or needs the full 1M context in an agent harness rather than a chat window. At $1.50/$7.50 in batch there is no quota to manage and no five-hour clock.
And treat the testimonial for what it is. Someone with good taste in models subscribed to Ultra, said so, and got 1,216 people nodding. The interesting fact underneath is smaller and more useful: the model they were praising was already on the cheaper plan, and it had been since the day it shipped.
How we checked this
Plan inclusions, prices and usage-limit wording come from Google One's plan comparison and Google's Gemini apps support article, both read on 21 November 2026; the AI Pro price of $19.99 and the Ultra price of $99.99 are the figures on those pages, and the $80 gap is arithmetic on them. Antigravity's model availability table, default-model statement and 512K context figure come from its own documentation, read the same day. The 1M-token API context, the $3.00/$15.00 rate card, the batch, flex and priority rates and the 18 November 2026 GA date come from Google's Gemini API changelog and pricing page. The 61.4 / 59.6 / 60.0 Intelligence Index figures, the index version v5.0.1 and the $6.00 blended price are Artificial Analysis's published page for the model, read on 19 November and labelled here as a third-party measurement rather than a Google claim. The post that prompted this piece is @kimmonismus on X, timestamped 20 November 2026 at 23:27 UTC; its like count is a reading, not a finding. Where Google publishes no figure — the 20x Ultra price — we say so rather than estimate it.
