The Claude Opus family
Opus is the top tier of the Claude architecture, and its unusual parameter is that the rate has held flat from Opus 4.5 through Opus 5: still $5 in and $25 out. The version that belongs in an active deployment today is Opus 5.
- fixed rate across five versions
- one-million-token context window
- the top tier of the Claude architecture
Last checked: This is a reference page. It is re-checked against the vendor sources and updated when a new version ships. Change log
Which version to deploy
Opus 5. The only technical justification for staying on 4.8 is a service that has tested and validated that exact version behaviour and does not want any variable disturbed. On unit cost there is no justification for holding off migration at all, since both versions carry the same rate.
A pattern that repeats across this family architecture
On this production line, Anthropic has held the rate fixed and raised functional capacity with each release. Opus 4.5, 4.6, 4.7, 4.8, and 5 are all logged at $5 and $25. For any service built on this family, that means migration between versions carries zero financial risk, with only behavioural risk remaining.
One significant exception exists in this pattern: from version 4.7 onward the tokenizer changed, and the same input text now produces roughly 30 percent more tokens. So the unit rate is flat while the real invoice can rise, and this parameter has to be measured separately from the nominal rate.
Release timeline
Each version with its own release date, and what changed against the one before it.
- Claude Opus 5 current Half the price tier of Fable 5 with near-equivalent throughput on coding workloads, and a knowledge cutoff ahead of the entire Anthropic model fleet.
- Claude Opus 4.1 retired
- Claude Opus 4.5 legacy
- Claude Opus 4.6 legacy
- Claude Opus 4.7 legacy
- Claude Opus 4.8 legacy
Known weaknesses and limits
- The rate held flat, but the tokenizer changed at version 4.7, so real token consumption has increased.
- Older versions in this family remain available, which is exactly how a deployment ends up sitting on a deprecated build unintentionally.
- No version in this family has an official deployment route from Iran.
Technical verdict
If you have a service deployed on this family, migrate to Opus 5 sooner rather than later, since the unit cost of that migration is zero. Just measure real token consumption before migrating, not only the nominal rate.