Anthropic's Newest AI Is Better and Cheaper — Here's the Catch
The same AI brainpower now costs less per task. Safety testing got thinner.
Anthropic is introducing Claude Opus 5.5 — the world's most powerful model, at least by some measures like Artificial Analysis or any standard benchmark list. Anthropic claims Opus 5.5 is outright as good or better than Fable 5.1, while being actively cheaper than Opus 5. Quick feedback from the internet is that Opus 5.5 is very good, but according to the article's author, more time is needed before offering comment. One big policy change: Anthropic will no longer test helpful-only versions of Claude, instead using tests designed to avoid refusals. According to the article, this is not a free change — evals that had refusal issues got dropped, and in one section multiple teams lost time to issues with refusals.
- Claude Opus 5.5 is Anthropic's most powerful AI yet, and it costs less than the model it replaces — good news for anyone paying per task
- Anthropic stopped testing a special version of the AI with no safety refusals, and some older safety checks were dropped as a result
- The safety report covers bioweapons, hacking and AI acting on its own; Anthropic says it stays below its danger lines, but reviewers want more time
Why It Matters
Cheaper, smarter AI reaches your everyday work tools sooner — but the safety testing behind it is getting less transparent.