Claude 4.5 is out there all over the place right now. Via the API, the mannequin maintains the identical pricing as Claude Sonnet 4, at $3 per million enter tokens and $15 per million output tokens. Builders can entry it by way of the Claude API utilizing “claude-sonnet-4-5” because the mannequin identifier.
Different new options
Some ancillary options of the Claude household obtained some upgrades right now, too. For instance, Anthropic added code execution and file creation straight inside conversations for customers of Claude’s internet interface and devoted apps. Alongside these strains, customers can now generate spreadsheets, slides, and paperwork with out leaving the chat interface.
The corporate additionally launched a five-day analysis preview referred to as “Think about with Claude” for Max subscribers, which demonstrates the mannequin producing software program in actual time. Anthropic describes it as “a enjoyable demonstration exhibiting what Claude Sonnet 4.5 can do” when mixed with acceptable infrastructure.
A screenshot of the accessible Anthropic AI fashions for Claude Max customers seen within the Claude internet interface on September 29, 2025.
Credit score:
Benj Edwards
As talked about above, the command-line improvement instrument Claude Code additionally acquired a number of updates right now, alongside the brand new mannequin. The corporate added checkpoints that save progress and permit customers to roll again to earlier states, refreshed the terminal interface, and shipped a local VS Code extension. The Claude API additionally positive aspects a brand new context modifying characteristic and reminiscence instrument for dealing with longer-running agent duties.
Proper now, AI corporations are significantly clinging to software program improvement benchmarks as proof of AI assistant functionality as a result of progress in different fields is tough to objectively measure, and it is a area the place LLMs have arguably proven excessive utility in comparison with different fields that may undergo from confabulations. However folks nonetheless use AI chatbots like Claude as common assistants. And given the latest information about troubles with some customers taking place fantasy rabbit holes with AI chatbots, it is maybe extra notable than normal that Anthropic claims that Claude Sonnet 4.5 reveals lowered “sycophancy, deception, power-seeking, and the tendency to encourage delusional pondering” in comparison with earlier fashions. Sycophancy, specifically, is the tendency for an AI mannequin to reward the person’s concepts, even when they’re incorrect or probably harmful.
We might quibble with how Anthropic frames a few of these AI output behaviors by way of a decidedly anthropomorphic lens, as we now have prior to now, however general, makes an attempt to scale back sycophancy are welcome information in a world that has been more and more turning to chatbots for excess of simply coding help.
