· via dev.to (home feed)
Gemini API standard prices set to double for Carbon and Argon in December, Flash in January
Google's pricing page schedules doubled standard rates for Gemini 4 Carbon and Argon from 1 December 2026 and for Gemini 3.6 and 3.8 Flash from 1 January 2027, with batch prices rising as well.

Standard rates double on two fixed dates
Google's Gemini pricing page carries scheduled increases that will double standard paid-tier prices for several models, according to a post on dev.to that recorded the page's figures on 18 November 2026. The changes take effect on fixed calendar dates rather than being tied to any change in usage, so bills move even when workloads stay exactly the same.
Per the figures in that post, with prices in USD per 1 million tokens at the standard paid tier:
- Gemini 4 Carbon — $2 input / $10 output now; $4 input / $20 output from 1 December 2026
- Gemini 4 Argon — $2 input / $10 output now; $4 input / $20 output from 1 December 2026
- Gemini 3.8 Flash — $0.75 input now; $1.50 input from 1 January 2027
- Gemini 3.6 Flash — $0.75 input now; $1.50 input from 1 January 2027
- Gemini 4 Hydrogen — $1.50 input, with no change listed on the page
The consistent pattern across the page, the post notes, is that these standard prices double.
Batch prices rise as well
Batch rates on the same page also double; Gemini 4 Carbon batch input, for instance, goes from $1 to $2 per million tokens. The relative discount for batching is preserved, but the absolute savings shrink, so teams relying on batch pricing absorb part of the increase too.
The arithmetic for an existing workload
The post walks through an example: a workload consuming 10 million input tokens and 2 million output tokens per month on Carbon or Argon.
Until 30 November 2026, that traffic costs 10 × $2 plus 2 × $10, or $40 per month. From 1 December 2026, the identical workload costs 10 × $4 plus 2 × $20, or $80 per month. Nothing about the application changed; only the calendar did.
What teams can do before the dates hit
The post offers several practical steps, paraphrased here:
- Recalculate cost estimates with the new rates now, ahead of the effective date, rather than discovering the change on the first invoice at the higher price.
- Trial a cheaper tier for tasks that do not require the largest model, and measure output quality on your own prompts rather than assuming the smaller model is good enough.
- Use batch and caching where the provider offers them. They still reduce costs, but batch prices are rising alongside standard rates.
- Compare vendors on their own published pricing pages before the increases land; the post points to Anthropic's and OpenAI's pricing documentation as reference points.
Why it matters
Price changes tied to fixed dates are a different kind of budgeting hazard than usage-driven cost growth. A team that sized a product around Carbon or Argon at $2 in / $10 out will see its inference bill double overnight on 1 December 2026 with no code change, and Flash users face the same a month later. Multi-month forecasts, unit economics and per-customer margins all shift at once.
The saving grace is that these increases are published in advance, which makes them avoidable surprises. Teams still have lead time to re-run forecasts, evaluate smaller models against their real workloads, and weigh competing providers on current pricing. One caveat is worth stating plainly: the figures above come from a third-party post that read Google's page on a single date and notes it was written with AI assistance. Anyone budgeting against these dates should verify Google's own pricing page directly before committing numbers.
- #gemini
- #api-pricing
- #llm
- #cost-management