The latest version of Claude, Opus 4.5, has been released and has outperformed all its competitors in programming.

It is the first model to achieve an 80% SWE-Bench score. The API price is $5/$25 per million tokens.

Developers can now choose how the model allocates tokens: quickly and cheaply, or slowly for maximum performance.

Claude Code has also been added to the desktop app. Long conversations should no longer be interrupted. When necessary, Claude will automatically extract the previous context and continue the chat.