Introducing Claude Opus 4.5
- Source
- anthropic.com
- Author
- Anthropic News
- Date
Context
Opus 4.5 introduced an effort setting that lets developers choose how much work the model puts into a task. Anthropic's November 24, 2025 announcement describes the control as a way to balance capability against time and spending. In its software-repair , the company reported that medium-effort Opus 4.5 matched Sonnet 4.5's best score while using 76% fewer output . That result illustrates why model choice and effort level should be evaluated together: a lower-effort run can be sufficient for some workloads. The percentage is a result, not a promised reduction in application costs.
- SWE-bench — The standard benchmark for AI coding agents: real GitHub issues from real repositories, scored by whether the agent's patch passes the project's own tests.
- eval — A repeatable test for AI quality — a set of tasks plus scoring — used the way software teams use test suites, because model output is too variable to judge by eyeballing.
- token — The chunk of text a model reads and writes in — roughly three-quarters of a word — and the unit AI usage is billed in.
- benchmark — A standard public test set for comparing AI models — the shared scoreboards behind every "model X beats model Y" claim.
“It’s intelligent, efficient, and the best model in the world for coding, agents, and computer use.”
“Opus 4.5 is a step forward in what AI systems can do, and a preview of larger changes to how work gets done.”
“Pricing is now $5/$25 per million tokens—making Opus-level capabilities accessible to even more users, teams, and enterprises.”
“Testers noted that Claude Opus 4.5 handles ambiguity and reasons about tradeoffs without hand-holding.”
“Early testing shows it surpasses internal coding benchmarks while cutting token usage in half , and is especially well-suited for tasks like code migration and code refactoring.”
Mario Rodriguez
Checking sign-in…
Loading comments…






