AI & ML 4 min read

Gemini 4 Hydrogen Costs $1.50 per Million Input Tokens. Argon and Carbon Cost 2.67 Times That From 1 December

Google's newest and lowest-priced model has one flat rate across its 1M-token context. We priced the same job on its two dearer siblings, before and after 30 November.

Maya Lin
Digital Platforms Analyst
Published 14 Nov 2026, 11:55 AM (SGT)
Share:
A red clamp squeezing a stack of coins on top of a calculator A red clamp squeezing a stack of coins on top of a calculator Photo by stevepb on Pixabay
Advertisement

14 NOV 2026 — Google released Gemini 4 Hydrogen on 12 November at $1.50 per million tokens of input and $7.50 per million of output, one flat rate across the model's whole 1 million-token context window. It is the lowest-priced of the three Gemini 4 models on Google's API price list. It is rolling out to all tiers of Google AI Studio and the Gemini API, the Gemini app and Google Search.

Google's launch post says that at its default "high" thinking level Hydrogen scores 51 on the Artificial Analysis Intelligence Index, a point above Gemini 4 Argon's 50, and 54 at a new "xhigh" setting. The four levels are low, medium, high and xhigh. Google says Hydrogen finishes tasks 2.5 times faster than Gemini 3.8 Flash and scores 57% on Terminal-Bench 4.1, against the older model's 18.2%. It is also adding Orchestrator to Google Antigravity, which switches between Carbon and Hydrogen inside one session; on Terminal-Bench 4.0, the company says, it came near Carbon's success rate while using 24% less quota on average and finishing 19% faster. These are Google's figures. We have not tested the model.

The gap widens on 1 December

Carbon, which Google's pricing page describes as its most capable reasoning model, and Argon are both currently listed at an introductory $2 and $10 per million tokens. Google's API pricing page lists both at $4 and $20 from 1 December 2026, and the Carbon announcement of 10 November gives the same end date for its introductory period. The pricing page shows no end date for Hydrogen's rate.

ModelInput, $ per 1M tokensOutput, $ per 1M tokensWhen the price applies
Gemini 4 Hydrogen1.507.50From launch on 12 Nov; no end date listed
Gemini 4 Argon2.0010.00Through 30 Nov 2026
Gemini 4 Argon4.0020.00From 1 Dec 2026
Gemini 4 Carbon2.0010.00Through 30 Nov 2026
Gemini 4 Carbon4.0020.00From 1 Dec 2026

Paid-tier standard rates. Output prices include thinking tokens.

On those list prices, Argon and Carbon cost 1.33 times as much as Hydrogen until the end of November ($2 against $1.50, and $10 against $7.50). From 1 December the multiple is 2.67 ($4 against $1.50, and $20 against $7.50).

Advertisement

Take one job that reads 1 million tokens and writes 200,000. On Hydrogen that is $1.50 for input plus 0.2 × $7.50 = $1.50 for output, or $3.00. The same token counts on Argon or Carbon come to $2.00 + $2.00 = $4.00 today, and $4.00 + $4.00 = $8.00 from December.

Equal token counts are an assumption

That sum assumes every model spends the same tokens on the task. Google's own posts say they do not. The Carbon post describes that model as "using more tokens than Argon to explore thoroughly". For Hydrogen, Google says partners saw token use fall by 12% to 23% compared with 3.8 Flash. Google says the thinking levels mean users "only wait and pay for the reasoning depth a task requires", and the pricing page bills thinking at the output rate, so the xhigh setting that lifts Hydrogen to 54 on Google's account also costs more per task. The real ratio depends on tokens per task, measured on a team's own work.

One difference needs no arithmetic. The pricing page offers Hydrogen on the API's free tier, where Argon and Carbon are listed as not available. We reported Argon's coming price change in October.

Advertisement
Maya Lin
Digital Platforms Analyst

Maya Lin covers SaaS platforms, workflow automation, creator tools, and productivity software for RECATOOLS.

View author profile → · Editorial policy

About this byline Maya Lin is a RECATOOLS editorial persona used for platform and productivity coverage. Articles are produced and reviewed under RECATOOLS editorial supervision.

Corrections policy

Advertisement