ModelBrief · Gemini 4 Neon · Topic Page
Last updated Nov 20, 2026

Gemini 4 Neon Pricing: What $3/$15 Means Before and After December 1

Published Nov 19, 2026 · Last updated Nov 19, 2026

Gemini 4 Neon's price list is a single line: $3 in / $15 out per 1M tokens, $0.15 cached input, no promotional price, no free tier. The complicated part is the calendar — Carbon and Argon's $2/$10 is a promotional price that ends November 30, rising to $4/$20 on December 1. The same price list gives opposite conclusions in November and December. Here are the three sums, then the cost-per-task figure where two sources disagree.

The price list: four tiers

API tier (per 1M tokens)InputOutput (incl. thinking tokens)Cached input
Standard$3.00$15.00$0.15
Batch API$1.50$7.50$0.075
Flex$1.50$7.50$0.075
Priority$5.40$27.00$0.27

Caching is implicit: it triggers automatically on prompts of 4,096 tokens or more, at a 95% discount; explicit caching is not supported. Batch and Flex are half the standard price; Priority is 1.8x standard. Output pricing includes thinking tokens — higher thinking tiers mean bigger bills, as the next sections show.

First sum: Neon costs 2x Hydrogen

Hydrogen's standard price is $1.50/$7.50, so Neon's $3/$15 is exactly double, cached input included ($0.15 vs $0.075). One useful equivalence: Neon's Batch / Flex price ($1.50/$7.50) equals Hydrogen's standard price — batch jobs that are not urgent get Neon's Batch tier at Hydrogen's standard rate, trading up from an AA index of 54 to 61. The other way around, Hydrogen has a free tier and Neon does not: zero-budget trials still mean Hydrogen.

Second sum: until November 30, Neon costs 50% more than Carbon

Carbon and Argon are currently on promotional pricing: $2 in / $10 out, ending November 30. Neon's $3/$15 is a full 50% higher on both input and output. A concrete example: 1M input tokens plus 1M output tokens costs $18 on Neon against $12 on Carbon — on launch day, using Neon costs $6 more than using the flagship. So "Neon is cheaper than Carbon" is not true per token in November; VentureBeat and The Decoder both made this point in their launch-day coverage.

Third sum: from December 1, Neon is 25% cheaper than Carbon

From December 1, Carbon and Argon rise to $4/$20, and Neon at $3/$15 becomes 25% cheaper on both input and output. The same 1M + 1M tokens: Neon still $18, Carbon now $24. Neon's pricing strategy only makes sense against this date: it is not "always cheaper than Carbon" but "cheaper than Carbon from December 1" — Google wrote the promotional expiry into the pricing page, and Neon's price is anchored against Carbon's post-December-1 rate. Whether $3/$15 is permanent, Google has not said; longer-term pricing after December 1 is currently unknown.

Cost per task: Google's claim vs LMArena's snapshot

Beyond per-token prices there is cost per task — what a model actually spends to finish one job. Google says Neon costs 32% less per task than Carbon at the default high tier, and 36%+ less than top competing models at their highest settings. That is Google's own measurement on AA index tasks: Neon spends fewer thinking tokens and gets things right the first time more often, so the total per task is lower even though the per-token price is higher.

But another snapshot points the other way: LMArena's Agent cost chart (via kingy.ai, Nov 18) shows a median $2.66 per task for Neon against $1.34 for Carbon — Neon twice as expensive. The methods differ: AA index tasks are standardized benchmark tasks, while Arena costs come from real user sessions; sample sizes, task mixes and thinking tiers all differ. They cannot be reconciled yet, and this site treats neither as settled. Notably, AA's own data supports doubting the "Neon is cheaper" story: one xhigh index run generates 283M output tokens, 2.3x the 123M of high — the top tier's score is bought with tokens.

The thinking tier is the cost tier

The most effective knob for Neon's cost is the thinking tier, not the output cap (as with Carbon, the output cap covers thinking tokens — set it too low and you get an empty answer back). Output tokens per AA index run: low 28.3M, medium 62.9M, high 123M, xhigh 283M. High uses less than half the tokens of xhigh and loses only 1 point (60 vs 61); medium uses half again and loses 5 points (55). Use high for everyday work and switch xhigh on only for the hardest jobs — that is the cheapest way to run Neon. Batch / Flex at half price is the second-best knob: non-urgent work on Batch halves the rate outright.

Subscriptions: Pro is the floor, no free tier

Using Neon in the Gemini app needs a Google AI Pro or Ultra subscription; Antigravity likewise needs a paid Pro / Ultra subscription. Free users cannot reach it at all. The Pro-vs-Ultra gap that exists for Carbon does not exist here — Carbon is Ultra-only, while Neon lowers the bar to Pro. But subscription prices and Neon's exact usage quotas were not verified in this round: Google's launch materials do not publish Neon quotas for Pro / Ultra plans, so this page states no subscription prices — Google's official subscription page is authoritative.

Our take: Neon's pricing story has two chapters: in November it is "the second tier that costs more than the flagship on promotion"; from December 1 it becomes "the mid-tier that is 25% cheaper than the flagship". Use it now with Batch / Flex at half price and the high thinking tier — the only two certain discounts. Whether a task really costs 32% less than Carbon or twice as much, the two available measurements disagree; wait for independent numbers before treating either as fact.

Reference sources: this article is an original rewrite produced by this site; facts and figures were checked against Gemini API pricing (read Nov 19, 2026), Google's official blog (Nov 18, 2026), Artificial Analysis' model page (read Nov 19, 2026) and kingy.ai's LMArena cost snapshot (Nov 18, 2026)

← Back to the Gemini 4 Neon topic page