Anthropic Tokens Fell 50%; 7 Gemini Prices Are Set to Double

Google’s Gemini API pricing page, last updated on 22 October 2026, lists a dated price step-up for seven of its 32 model listings, and every one of those steps doubles the price. Gemini 4 Argon’s output price is listed at “$10.00 through November 30, 2026. $20.00 starting December 1, 2026.” The same week, the data firm YipitData wrote that “across our token-based datasets, we’ve observed a ~50% Y/Y decline in token prices.”

The two figures measure different things. Google’s are list prices with dates attached; YipitData’s is a measured change across its own panels, in a note on Anthropic. This piece sets them side by side, with the other prices from our October reporting, and does not combine them.

Business Pill · COST PER TASK

A one-minute explainer of cost per task: pricing the job rather than the word. It teaches the general idea only and says nothing about any company in this story.

The Schedule, Model by Model

A token is a small piece of text, roughly a short word, and AI providers price their models per million tokens, with separate prices for what goes in (input) and what comes out (output). The rows below are the standard paid tier on Google’s page, read on 24 October 2026.

Per 1 million tokens, US dollars, standard paid tier. Output prices for the text models include thinking tokens; the Robotics ER 2 input price covers text, image, video and audio. Source: Google, Gemini Developer API pricing page, last updated 22 October 2026, read 24 October 2026.

Gemini 4 Argon’s other tiers follow the same pattern on the page: the priority tier is listed at $3.60 input and $18.00 output through 30 November, then $7.20 and $36.00 from 1 December. By our arithmetic, every scheduled step in the table is exactly two times the current price.

What Google Calls the Current Prices

For Gemini 4 Argon, Google has described today’s rate as introductory. A footnote on Google’s blog of 13 October, quoted in our Argon launch piece, reads: “After the introductory period expires on November 30, 2026, the price of $4 per 1M input tokens and $20 per 1M output tokens will apply.” Our earlier Argon piece reported an earlier version of the footnote on Google’s announcement of 30 September, giving the same $4 and $20 prices but no end date.

So the Argon step-up is not a new price rise announced this week: it is the end of an introductory period Google has described since September. For the other six listings, the page gives the step-ups as plain dated prices and does not state a reason. What the page adds is the breadth: the Flash, speech and robotics listings carry the same doubling at the turn of the year.

Two Measures, Not One

YipitData’s ~50% figure comes from its Anthropic Pre-IPO Overview of 22 October, reported in our token-price piece. It lists the contributors: “Price cuts from Opus 4.5, Sonnet 5’s introductory pricing that became permanent, greater cache hit rates, and Fable 5.1 cache hit pricing reductions have contributed.”

One line in that list, according to YipitData, describes the opposite path to Argon’s: an introductory price that became permanent. For Argon, Google’s page dates the end of the introductory price instead. Neither source says how the other company’s customers will respond, and YipitData’s decline is measured across its own datasets, not on any single list price.

The same piece embeds a post by Puru Saxena, founder of AlphaTarget, who says September AI token dollar spend was “basically flat (+0.2% MoM)” while GPU rental prices kept rising. Citing Vercel usage, he also says the average token price fell about 47%. Those are his figures.

Connecting the Coverage

Other prices from our October reporting give a range. Moonshot listed Kimi K3.1 Pro at $2.00 input and $10.00 output, which our Kimi piece found to be half of the $4 and $20 Anthropic lists for Claude Opus 5.5. That is the same as Gemini 4 Argon’s price today and half of its price from 1 December, by our arithmetic.

Perplexity lists its decision model at $0.01 per million input tokens on its hosted API and runs it free on a Mac, our Perplexity piece reported; OpenAI lists its Decisions API at $0.10, according to our Decisions API piece.

On the supply side, “2027 is largely sold out,” Amazon chief executive Andy Jassy said of AWS computing capacity, Reuters reported on 23 October, as covered in our AWS capacity piece.

The Structural Read

As we read it, the only step-up Google has tied to an introductory period is Gemini 4 Argon’s; the page gives the other six as dated prices without a stated reason.

As we read it, the paths differ: YipitData lists an Anthropic introductory price that became permanent among the contributors to its decline, while Google’s page dates the end of Argon’s.

As we read it, the low end of our October reporting sits elsewhere: Perplexity at $0.01 per million input tokens and OpenAI’s Decisions API at $0.10 per million input tokens, prices for narrower decision models.

Three Implications

FOR TEAMS BUDGETING GEMINI USAGE Google’s page lists Argon at $10.00 output through 30 November 2026 and $20.00 starting 1 December 2026, and the Flash listings step up on 1 January 2027.

FOR READERS OF TOKEN-PRICE DATA YipitData says its ~50% decline is measured across its token-based datasets, and Puru Saxena said September AI token dollar spend was basically flat (+0.2% MoM) and, citing Vercel usage, that the average token price fell about 47%.

FOR RIVAL PRICING Moonshot lists Kimi K3.1 Pro at $2.00 input and $10.00 output, which by our arithmetic equals Argon’s price today and half of its December price.

The Business Engineer Lens

This story maps onto the Business Engineer framework The State of The Inference Economy.

The framework puts it as “Inference is a revenue engine: continuous, distributed across millions of applications and agents, incurring marginal cost with every query.”

As we read it, the list price per token is where that marginal cost meets the customer, and Google’s page now dates seven of those prices. The framework was published in March 2026 and does not discuss these figures.

What Is Not Established

The sources do not say whether Google will extend Argon’s introductory period, why the six other listings step up, whether customers will move to other models, or what share of Gemini usage runs at these list prices rather than at negotiated or discounted rates. Google’s page lists other models without a dated step-up. YipitData does not give a figure for Google, and Saxena’s numbers are his own reading of the data he cites.

The Bottom Line

YipitData measured token prices falling about half in a year and, according to YipitData, one Anthropic introductory price became permanent along the way. Google’s own page schedules the opposite for seven listings: each doubles between 1 December 2026 and 1 January 2027, with Argon’s step marking the end of its introductory period.