28 October 2026 · Site Methodology & Pricing Notes · 6 min read
I went to add Claude Sonnet 5 to our model-comparison list and found it wasn't there at all -- not stale, not outdated, just absent. Neither was Sonnet 5.5, Opus 5, Opus 5.5, or Haiku 5.5. Our shared model-pricing file still stopped at Opus 4.8 and Sonnet 4.6, like the last two Anthropic release cycles never happened. Pulling the current numbers to fix it, I found something odder than a missing row: a price increase Anthropic told everyone was coming, on a specific date, that it turns out never landed.
When Sonnet 5 launched, Anthropic priced it at $2/MTok input and $10/MTok output -- and said so explicitly: introductory pricing, in effect through August 31, 2026, with a bump to $3/$15 per million tokens scheduled for September 1. That's a real date, publicly committed to, the kind of thing you'd budget around if you were running a high-volume Sonnet integration. Anthropic's pricing page now carries this exact footnote, current as of today: "The $2/$10 per million input/output token pricing for Claude Sonnet 5, announced at launch as introductory pricing through August 31, 2026, is now the standard price. The previously scheduled increase to $3/$15 per million input/output tokens on September 1, 2026 will not occur."
No blog post, no announcement I could find -- just a footnote correcting the record. The hike was real enough to get a specific date attached to it, and specific enough to get walked back in writing once it quietly didn't happen.
| Modelo | Entrada | Saída |
|---|---|---|
| Claude Sonnet 5.5 | $2/MTok | $10/MTok |
| Claude Sonnet 5 | $2/MTok | $10/MTok |
| Claude Sonnet 4.6 (previous gen) | $3/MTok | $15/MTok |
| Claude Opus 5.5 | $4/MTok | $20/MTok |
| Claude Opus 5 | $5/MTok | $25/MTok |
| Claude Opus 4.8 / 4.7 / 4.6 / 4.5 (previous gens) | $5/MTok | $25/MTok |
Sonnet 5 and 5.5 are 33% cheaper per token than the Sonnet 4.6 generation they replaced, on both input and output. Opus 5.5 is 20% cheaper than every Opus version back through 4.5. That's backwards from the usual pattern, where a newer, stronger model either holds price or costs more -- GPT-5.5 at $5/$30 versus GPT-5 at $1.25/$10 is the more familiar direction, newer and pricier. Anthropic's two most recent releases both moved the other way.
Our shared model list -- the data file behind the main calculator, the cheapest LLM API finder, and roughly thirty other pages -- had thirteen Anthropic entries, topping out at Opus 4.8 and Sonnet 4.6. Not one of the five models above existed in it. Anyone running our "cheapest LLM API" comparison for a Sonnet-class workload was being shown 4.6's $3/$15 as Anthropic's best offer, when the actual best offer was 33% lower and had been for a while.
Fixed by adding all five current entries -- Opus 5.5, Opus 5, Sonnet 5.5, Sonnet 5, Haiku 5.5 -- at their verified current prices to the shared file. Because that file is one script loaded by every page in every language, it's a single edit, not thirty.
I default to assuming AI API prices only move up. That assumption is why the gap sat there long enough to need its own blog post -- I wasn't rechecking Anthropic's page because I wasn't expecting good news. The lesson isn't "always expect a discount." It's that a provider cancelling a scheduled hike doesn't get announced the way the hike itself would have been. If you want to know the real number, you go read the footnote, not the headline you remember from launch day.
New to usage-based API pricing in general? Start with the guias gratuitos de custo de API.
Implemente você mesmo: DigitalOcean – crédito grátis de $ 200 ↗ · VPS Hostinger ↗
Checked directly against platform.claude.com/docs/en/about-claude/pricing, 2026-10-28. Reference figures, not live quotes -- confirm current terms on Anthropic's pricing page before budgeting.