SMRTR summary
Claude Opus 4.8 outperformed Opus 4.7 in honesty tests, better avoiding fabricated citations and overconfident diagnoses — but failed by incorrectly assuming a user's location applied to his father's insurance claim.
SMRTR provides this summary for quick context. The original article belongs to Reddit.
Read the original articleThis archive is built from SMRTR newsletter stories. Subscribe for hand-picked stories without the extra noise.
Anthropic released Claude Opus 4.8, which is four times more likely to flag its own coding errors than its predecessor, at unchanged pricing. More significantly, Mythos-class...
Anthropic released Claude Opus 4.7, an iterative upgrade over Opus 4.6 that falls well short of the more powerful Claude Mythos. While safety and cyber capabilities remain largely...
Anthropic released Claude Opus 4.5 just one week after Google's Gemini 3 Pro claimed top AI performance rankings, delivering significant improvements in specialized areas despite...
Anthropic released Claude Opus 4.8 at the same price as its predecessor — $5 per million input tokens and $25 per million output tokens — while delivering measurable performance...
Claude Opus 4 was used to apply Bootstrap styling to a Rails CRUD app. The AI analyzed the project, understood its structure, and made targeted improvements to the views. It...
Anthropic launched Claude Opus 4.5, a new AI model that significantly outperforms previous versions in coding, software engineering, and computer tasks while using fewer tokens...