The top-ranked AI model's price doubles on 1 December
Google's own pricing page lists Gemini 4 Argon, the model at the top of the LMArena text leaderboard when we checked, at $2.00 input and $10.00 output per million tokens through 30 November, then $4.00 and $20.00. We read the pricing pages of Google, OpenAI, Anthropic, DeepSeek, xAI and Mistral on 26 October 2026, archived each one, and list every dated change they state. Three other dated changes are on the calendar, the last on 1 January 2027.
In this article
The calendar: four dated changes, soonest first
Each item below is a statement on the provider's own page, with the date it gives. Nothing here is inferred from a third-party price list. Days are counted from 26 October 2026, the day we read the pages.
Calendar: subscribe in your calendar app or download the .ics file. It lists these changes and the model retirements on our retirements page; events marked tentative are earliest-possible dates, not announcements.
-
1 November6 days after we read the pagesNew feeDeepSeek
Search and fetch stop being free
Server-side search and fetch calls made through DeepSeek Harness, Claude Code or Codex are free today. DeepSeek's page says billing starts at 00:00 Beijing time on 1 November 2026, at $0.75 per 1,000 successful requests.
Source: DeepSeek Models & Pricing (full DeepSeek breakdown), read 26 October 2026
-
21 November26 days after we read the pagesPromotion endsOpenAI
GPT-5.6 Sol promotional pricing
OpenAI's pricing page says GPT-5.6 Sol promotional pricing is available at least through 21 November 2026. The page does not say what the price becomes afterwards.
Source: OpenAI API pricing, read 26 October 2026
-
1 December36 days after we read the pagesPrice doublesGoogle
Gemini 4 Argon, the top-ranked model
Paid tier: $2.00 input and $10.00 output per million tokens through 30 November, then $4.00 and $20.00 from 1 December.
Source: Google Gemini API pricing, read 26 October 2026
-
1 January 202767 days after we read the pagesPrice doublesGoogle
Gemini 3.8 Flash and Gemini 3.6 Flash
Paid tier: $0.75 input and $3.75 output through 31 December, then $1.50 and $7.50 from 1 January 2027. Both models carry the same schedule.
Source: Google Gemini API pricing, read 26 October 2026
What it costs: a worked example
List prices are per million tokens, so the effect depends on your usage. As an illustration with our own assumption of 1,000,000 input tokens and 250,000 output tokens in a month:
| Model | Now | After the change | Change |
|---|---|---|---|
| Gemini 4 Argon, from 1 December | $4.50 | $9.00 | 2x |
| Gemini 3.8 Flash, from 1 January 2027 | $1.69 | $3.38 | 2x |
Assumption, not a measurement: 1,000,000 input and 250,000 output tokens per month. Standard paid tier, no caching or batch.
Best value right now: rating against price
We joined the LMArena overall text rating (published 22 October 2026, CC BY 4.0) to current list prices. The leader is Gemini 4 Argon at a rating of 1526. The cheapest model within 50 rating points of it is MiMo-V2.6-Pro at $0.54 blended per million tokens, 7.4x cheaper than the leader's $4.00 today and 14.7x cheaper after 1 December when the leader's blended price reaches $8.00.
Bars use a log scale. Blended = (3 x input + output) / 4. Models shown are those no other model beats on both price and rating. Rating differences inside a model's confidence interval are not meaningful.
The multipliers behind a list price
The number on a pricing page is a starting point. These are the multipliers the providers themselves state.
Fast costs 2x the standard input and 2x the output price on gpt-6-astra and gpt-6.1-sol. The page lists the same multiple for gpt-6-luna.
gpt-6-astra standard is $10.00 in and $50.00 out; Ultrafast is $60.00 and $300.00. gpt-6-luna has no Ultrafast row on the page.
Batch is half the standard price for the three flagship models listed.
DeepSeek says off-peak rates are half the peak rates. Peak is 01:00 to 04:00 and 06:00 to 10:00 UTC on weekdays, excluding Chinese public holidays.
Anthropic says Claude 4.7 and later models produce about 30% more tokens for the same text. Per-token list price for Opus 4.6 and 4.7 is the same, so the same text costs roughly that much more. Our estimate from their statement, not a measured bill.
OpenAI says regional processing (data residency) endpoints carry a 10% uplift for models released on or after 5 March 2026. FedRAMP endpoints carry one too.
Monthly plans: the first rung above $20 starts near $100
Consumer plans as displayed on 26 October 2026 for a United States visitor. ChatGPT Pro starts at $100, Claude Max at $100, and Google AI Ultra at $99.99 (a $199.99 Ultra tier is also listed, described as 20x the usage of AI Pro). The products are not equivalent; the pricing step is.
ChatGPT's page says the Go plan may include ads. Prices were read from the rendered pages on 26 October 2026 and are not archived as raw HTML. Providers state usage in relative terms (expanded, up to 5x), so no absolute limits are recorded. Perplexity, SuperGrok and Microsoft Copilot plans were not read.
How we know, and what we did not verify
Every dated claim above was read from a provider's own page on 26 October 2026. Static pages were saved with a SHA-256 hash; pages that build their price table with JavaScript were read in a browser and their rendered text saved the same way.
| Page | Read as | Date | SHA-256 (first 16) |
|---|---|---|---|
| ai.google.dev/gemini-api/docs/pricing | static HTML | 26 October 2026 | a252aac2059e41da |
| developers.openai.com/api/docs/pricing | static HTML | 26 October 2026 | cf734fcb60d01a59 |
| platform.claude.com/docs/en/about-claude/pricing | static HTML | 26 October 2026 | cc4da327e1278536 |
| api-docs.deepseek.com/quick_start/pricing | Chrome rendered text | 26 October 2026 | dfb3d0a8006d276f |
| docs.x.ai/developers/models | Chrome rendered text | 26 October 2026 | 9cd95f937ad43e71 |
| gemini.google/subscriptions/ | Chrome rendered text (verbatim excerpt) | 26 October 2026 | 24d8baff3cebb3e7 |
| platform.claude.com/docs/en/agents-and-tools/tool-use/web-search-tool | static HTML | 26 October 2026 | 77bf9b3190e190de |
| platform.claude.com/docs/en/about-claude/model-deprecations | static HTML | 26 October 2026 | 302f8bbb922bccea |
| developers.openai.com/api/docs/deprecations | static HTML | 26 October 2026 | 48c8cd40f4641e5b |
Corrections made while checking
- A third-party price list recorded xAI's Grok 4.1 Fast models rising to $1.25 input in September. xAI's current page lists a single flagship, Grok 4.7, at $2.00 input and $6.00 output, so we did not publish those as increases.
- The same list recorded DeepSeek V4 Flash rising. DeepSeek's page shows peak and off-peak rates, and the recorded figure matches the peak rate for a model name DeepSeek says is retired. That is not a clean list-price rise, so it is not published as one.
Limits
- Only the six providers named were read. A provider not listed may have scheduled changes we have not seen.
- A page can change after we read it. The hashes identify what we read.
- OpenAI's page does not say which row carries the promotional price, so we state no price for GPT-5.6 Sol.
- We have not asked any named provider to comment before publication. A provider or reader who finds a wrong figure can tell us; we correct it and date the correction.
Data files for this page are published with the AI Price Watch dataset on nesyona.com under CC BY 4.0. The LMArena rating data is CC BY 4.0, attributed to LMArena.
FAQ
When does Gemini 4 Argon's price change?
Google's Gemini API pricing page, read on 26 October 2026, lists the paid tier at $2.00 input and $10.00 output per million tokens through 30 November, then $4.00 and $20.00 from 1 December. That is a 2x increase on both. The page describes the lower rate as available through that date, and we have not seen Google call it a promotion.
What does the doubling cost in a real month?
It depends on your usage. As a worked example with our own assumption of 1,000,000 input tokens and 250,000 output tokens in a month, Gemini 4 Argon costs $4.50 at the current rate and $9.00 from 1 December. The same usage on Gemini 3.8 Flash goes from $1.69 to $3.38 on 1 January 2027.
Is DeepSeek search free?
Today, calls made through DeepSeek Harness, Claude Code or Codex are free according to DeepSeek's page. The page says per-request billing starts at 00:00 Beijing time on 1 November 2026, at $0.75 per 1,000 successful requests for each of Search and Fetch.
What will GPT-5.6 Sol cost after the promotion?
OpenAI's page says the promotional pricing is available at least through 21 November 2026 and does not state a later price. We will not guess one. The page is re-read weekly and this answer changes only when OpenAI states it.
Are these list prices what I will pay?
They are the prices the providers print for standard use. Fast and Ultrafast speed tiers, batch discounts, regional endpoints, caching, peak hours and tokenizer changes all move what a task costs, and each is covered above where the provider states it.