Anthropic Releases Claude Opus 4.5 and Cuts Its Price by Two-Thirds

The new model leads on software-engineering tests and costs a third of what the previous top model did — part of a pattern across the industry in which each generation is both better and dramatically cheaper.

Anthropic released Claude Opus 4.5 today, six days after Google's Gemini 3 and in the middle of the busiest few weeks of model releases the field has had. It leads on the tests that measure real software engineering — fixing actual bugs in actual codebases — and on long-running tasks where a model has to keep working toward a goal across many steps. It also costs about a third of what the company's previous top model did.

The price is the part worth noticing. The story people tell about AI is that it gets more capable; the story that matters more to whether it spreads is that it gets cheaper, fast. A task that cost a dollar to run last year costs pennies now, across every lab. That is why AI keeps turning up inside other products: not because it became possible, but because it became affordable enough to put in front of everyone.

The competitive picture is a three-way tie that changes hands every few weeks. Google took the benchmark lead last Tuesday; Anthropic takes much of it back today; OpenAI will presumably answer. For anyone outside the industry this scoreboard matters less than it appears to. The practical fact is that three companies now sell roughly equivalent, very capable systems, and are cutting prices against each other — which is the condition under which a technology stops being a product and becomes infrastructure.