Anthropic has just announced the latest artificial intelligence model named Claude Opus 4. 8. According to the company, this is a significantly upgraded version in terms of user collaboration capabilities, especially in fields such as automation programming, multidisciplinary reasoning, AI computer control, knowledge work processing and financial analysis.
Test teams rate Opus 4.8 as more reliable and capable of making sharper judgments when performing automated tasks. The model is also improved in terms of fidelity, regularly warning when uncertain about the result instead of making unfounded claims.
According to Anthropic, internal reviews show that Opus 4.8 has significantly better ability to detect and recognize errors in its own code. The rate of missing errors without warning has decreased about 4 times compared to its predecessor.
Tests on safety and user aptitude also recorded positive results. Opus 4.8 scored higher in criteria such as respecting user autonomy and user-serving actions.
Anthropic also said that the fast response mode of Opus 4.8 has a processing speed 2.5 times faster than before, while operating costs are reduced by about 3 times.
