Anthropic Unveils Claude Opus 4.8, a New, More Powerful Model

Anthropic Unveils Claude Opus 4.8, a New, More Powerful Model

https://www.nytimes.com/2026/05/28/technology/anthropic-claude-opus-48.html

Publish Date: 2026-05-28 13:00:00

Source Domain: www.nytimes.com

Here is a polite and respectful summary of the provided article:

– Anthropic has unveiled a new AI model, the Claude Opus 4.8, which has shown significant improvements over its predecessor, particularly in generating computer code.
– Independent benchmark tests from Vals AI reveal that Opus 4.8 outperforms all other publicly available AI technologies in “vibe coding,” where AI creates software based on conversational English prompts.
– The new model demonstrated a 10 percent improvement over its predecessor on the vibe coding benchmark and also excelled in mathematics.
– Opus 4.8 shows superior performance as an AI agent by automating tasks using other software programs.
– The advancements in AI systems in coding and identifying software vulnerabilities are rapidly advancing the field of cybersecurity.
– Anthropic previously released a more powerful technology, Claude Mythos, which it deemed too dangerous for public release but shared with a small select group due to its potential to identify security vulnerabilities.
– Mythos is anticipated to eventually be shared with a broader sector, including government agencies, according to Anthropic.