Changelog
Changelog: what changed, when, and on whose word
Model releases, price changes and every recheck that touched a number on this site — newest first, each with the date it happened and the source behind it.
Every entry below records a change to the numbers on this site — a price added, a page added, a source rechecked. Newest first. A date here is the date the change was made, not the date a vendor announced something, unless the entry says so.
2026-11-20
Added Gemini 4 Carbon and GPT-6.2 Sol to the price matrix
Both models now sit in the matrix with their own rows and their own verification dates. Carbon carries two rows because Google publishes two rates: an introductory $2.00 input / $10.00 output, and $4.00 / $20.00 after the introductory period expires. GPT-6.2 Sol is listed at $2.00 input, $0.10 cached input and $10.00 output.
Sources: Google's launch post Introducing Gemini 4 Carbon, published 2026-10-28; OpenAI's API pricing page as reported by BetaNews. Both checked 2026-11-20.
Published the Carbon comparison page
Neon vs Gemini 4 Carbon shows what the same 10M-input / 2M-output workload costs on each side of Carbon's introductory period, and records that the price order flips when the period ends. The page does not give the end date, because Google's launch post does not.
First version of the site
Six content pages — home with the monthly cost calculator, price matrix, benchmarks, two comparisons, this changelog — and four pages about the site itself: about, contact, privacy, terms. Every figure on the site states its source and the date it was checked; where a vendor has published nothing, the cell reads not published.
2026-11-18
Checked Gemini 4 Neon and Claude Opus 5.5 rates
Gemini 4 Neon: $3.00 per million input tokens, $0.15 cached input, $15.00 output, 1.0M context. Claude Opus 5.5: $4.00 input, $20.00 output, fast tier at $8.00 / $40.00. No cached-input rate and no context window are published for Opus 5.5 in the sources checked, so both cells read not published.
Sources: llm-stats.com for the Neon row; Anthropic's published rates as reported by VentureBeat for the Opus rows. The full source table is on the price matrix.
Checked the two benchmark figures
Artificial Analysis Intelligence Index value of 61 for Gemini 4 Neon (xhigh), third overall, and Google's claim of 16 wins in 19 tests. The two are recorded in separate columns on the benchmark page and are never combined.
2026-10-28
Gemini 4 Carbon announced by Google
Google published Gemini 4 Carbon: our next frontier model, stating introductory pricing of $2.00 per million input tokens and $10.00 per million output tokens, the same introductory price as Gemini 4 Argon, with cached input tokens priced at 95% off the input token price. The launch post's footnote states that $4.00 input and $20.00 output apply after the introductory period expires, without giving the date.
This entry records what Google published on that date. The rates entered this site's matrix on 2026-11-20.
How this log is kept
- One entry per change, with the date the change was made and the source that caused it.
- Vendor announcements are logged as announcements. A launch date is not the date this site started publishing the model's prices, and the log says both.
- Corrections are new entries. When a rate we published turns out to be wrong, the correction appears as its own dated entry rather than an edit in place, so a reader who quoted the old figure can see when it changed.
- A recheck is an event. “Checked on 2026-11-18” means a person read the source that day; it is not a claim that the number stayed the same afterwards.
What this log does not contain
Editorial notes, design changes and anything without a number attached. If you are looking for a rate rather than a date, start at the price matrix or run your own volume through the calculator. The two benchmark figures have their own page: benchmark sources.
Prices and benchmarks verified 2026-11-18; Carbon and Sol rows verified 2026-11-20. ModelCost is not affiliated with Google or Anthropic.