Opus 4.8 AI: Faster, Cheaper, and More Honest?

Summary

A new AI model, Opus 4.8, has been released, promising improvements over its predecessor. Key updates include a significant price reduction for fast mode, making it three times cheaper. The model also aims to be more honest and avoid making unsubstantiated claims, a common issue in previous AI versions, as highlighted by an example where an AI failed to detect a user generating content without credits. Opus 4.8 also introduces dynamic workflows for handling larger tasks and potentially running hundreds of parallel sub-agents, though this feature is primarily beneficial for enterprise users or API developers. Effort control is another new option, allowing users to reduce usage by making the AI "think less." Early benchmarks show a marked improvement in quality compared to Opus 4.7, with impressive results in website generation tasks, including better SEO, SVG rendering, and handling of multilingual content. However, the reviewer notes that many of these capabilities were already present in Opus 4.7, suggesting that while Opus 4.8 is an improvement, the core technology is mature enough to build complex applications with existing models, provided users know how to leverage them. The pricing for standard usage remains unchanged, with the fast mode being the primary cost adjustment.