Introducing Claude Opus 4.8
Blog post from Anthropic
Claude Opus 4.8, the latest version from Anthropic, offers enhancements over its predecessor, Opus 4.7, with significant improvements in benchmarks, coding capabilities, and agentic tasks, while maintaining the same pricing. It introduces new features such as user-controlled effort levels and dynamic workflows in Claude Code, allowing the model to handle large-scale problems more efficiently. Fast mode is now three times cheaper, working at 2.5 times the speed of previous models. The updated model demonstrates enhanced reliability, reasoning quality, and tool efficiency, setting a new standard for enterprise AI by improving on practical tasks like legal work and financial document processing. Opus 4.8 also shows a marked improvement in honesty and alignment, with reduced rates of misaligned behaviors. The release includes updates like the Messages API and increased rate limits to accommodate higher token usage, aiming to deliver a better balance between quality and user experience. Anthropic plans to introduce even more advanced models as part of Project Glasswing, focusing on cybersecurity and enhanced intelligence.