Thursday, November 19, 2026
News

Is Gemini 4 Neon cheaper than Carbon for your workload?

Google launched Neon on November 18 between Hydrogen and Carbon. The comparison that decides a December bill is not Neon against today's Carbon rate, but Neon against the rate Carbon reverts to on the first.

Blue and red neon lights

Not this month. Neon lists at $3 per million input tokens on the Gemini API and $15 per million output, while Carbon runs at $2 and $10 through November 30. On December 1 Carbon goes to $4 and $20, which puts Neon 25 percent below it. The switch date, not the launch, is what changes a bill.

What does Gemini 4 Neon cost per million tokens?

$3.00 input and $15.00 output on the Gemini Developer API, paid tier only. Cached input is $0.15, implicit caching only, which Google's launch post calls a 95 percent discount. Those prices carry no date qualifier, unlike the Carbon and Argon rows on the same page, and no Gemini 4 row on that page tiers by prompt length.

The model string is gemini-4-neon. Google's model guide gives it a unified 1,048,576-token limit covering input and output together, a single-segment output cap of 65,536 tokens, which is double Argon's non-streaming cap, and four thinking levels, LOW, MEDIUM, HIGH and XHIGH, with HIGH as the default rather than MEDIUM. Explicit context caching is not supported, and thinking_budget returns an HTTP 400.

Is Neon cheaper than Carbon right now?

Not in November, which is the part worth working out before you switch. The pricing page sets Carbon input at "$2.00 through November 30, 2026. $4.00 starting December 1, 2026." and output at "$10.00 through November 30, 2026. $20.00 starting December 1, 2026." Argon carries the identical dated pair. So a job running today is cheaper on Carbon than on Neon, and the same job is cheaper on Neon from the first.

Put a workload through it. Take 10 million input tokens and 2 million output tokens in a month. On Neon that is $30 plus $30, so $60. On Carbon in November it is $20 plus $20, so $40. On Carbon from December 1 it is $40 plus $40, so $80. The same work costs half as much again on Neon this month and a quarter less from next month. Google's own claim is narrower and about efficiency rather than list price: a 32 percent lower cost per task than Carbon at the default setting. VentureBeat, citing third-party trackers, reports the gap narrowing to about 17 percent when Neon runs at XHIGH against Carbon at HIGH. We worked the November side of this in the Carbon price cliff.

What does Neon cost on Vertex AI?

The same $3.00 and $15.00, but only on the global endpoint. Non-global endpoints are 10 percent higher, $3.30 and $16.50. The batch tier is exactly half standard, $1.50 and $7.50, and the priority tier is 1.8 times standard at $5.40 input. Cache reads are $0.15 globally, and the page says cache write pricing is "effective January 12, 2027".

Vertex also describes Carbon's sub-December rate differently from the Gemini API page. Both pages print the dated rate, but Vertex adds a footnote the API page does not: the rate arrives as promotional pricing, delivered as 50 percent credits back against net spend on selected models in a period, and it does not apply to the Cyber Permissive variants. The effective cost lands in the same place, the mechanism does not, which matters if your finance team reconciles credits rather than rates. The same split showed up when we priced Argon on Google Cloud.

What do you give up by moving off Carbon?

Capability at the top end, on Google's own framing: Carbon "tackles your hardest, long-horizon projects" while Hydrogen "handles fast everyday execution", and Neon sits between them. Google's model guide names Neon the target upgrade for Gemini 3.1 Pro and for general coding and enterprise reasoning on Gemini 4 Argon, which stops short of 9to5Google's flat claim that Neon replaces Argon.

Availability differs too. Neon is rolling out through the Gemini API inside Google AI Studio, in Gemini for business, in Google Antigravity for paid AI Pro and Ultra subscribers, and in the Gemini app for Pro and Ultra. VentureBeat notes that in the app, Pro users can select Neon while Carbon needs Ultra. Neon is not on the free tier, and nor are Carbon or Argon. Hydrogen, at $1.50 and $7.50, is the only one of the four with free-tier access, and free-tier content there is used to improve Google's products while paid content is not.

What should IT teams do about Gemini 4 Neon today?

  • Run your real token mix through both rates before December 1, not the headline price: the answer flips on that date.
  • If you are on Carbon or Argon and your spend is mostly input, model the December 1 doubling now, because it applies to both.
  • Check whether your Vertex deployment calls the global endpoint. Non-global costs 10 percent more for the same model.
  • Move anything that tolerates latency to the batch tier, which is half price on Vertex.
  • If you reconcile Google spend through credits, ask your account team which mechanism applies to you, the dated list price or the credits-back promotion.
  • Pin thinking_level deliberately. Neon defaults to HIGH, where Argon defaulted to MEDIUM, so a straight model swap raises output token counts.

Sources