Model family

The Claude Opus family

Opus is Claude upper tier, and the odd thing about it is that the price has not moved from Opus 4.5 through Opus 5: still $5 in and $25 out. The version to use today is Opus 5.

  • one price across five versions
  • 1M token context window
  • Claude upper tier

current Maker: Anthropic

Last checked: This is a reference page. It is re-checked against the vendor sources and updated when a new version ships. What changed

Which version to use

Opus 5. The only reason to stay on 4.8 is a service whose behaviour you have tested and do not want disturbed. On unit cost there is no reason at all, because both charge the same.

A pattern that repeats in this family

On this line Anthropic has held the price and raised the capability. Opus 4.5, 4.6, 4.7, 4.8 and 5 are all $5 and $25. For anyone who has built a service on the family, that means moving between versions carries no financial risk, only behavioural risk.

One important exception: from 4.7 onward the tokenizer changed, and the same text produces roughly 30 percent more tokens. So the unit price is flat while the real invoice can rise, and that has to be measured separately.

Specifications

Context window 1 million tokens Source
Input price $5 per million tokens Source
Output price $25 per million tokens Source

Release timeline

  1. Claude Opus 5 current Half the price of Fable 5 with close to its coding performance, and the latest knowledge cutoff of any Anthropic model.

Where it falls short

  • The price held, but the tokenizer changed at 4.7, so real token consumption went up.
  • Older versions in this family are still available, which is exactly how a team ends up sitting on a stale one.
  • No version in this family has an official route from Iran.

Our take

If you run a service on this family, move to Opus 5 sooner rather than later, because the unit cost of doing so is zero. Just measure token consumption first, not only the price.

Sources

  1. Anthropic model overview vendor source 12 August 2026