Claude Opus 4.8

Claude Opus 4.8 adds multimodal and tooling capabilities to baseline

Claude Opus 4.8 expands significantly beyond its previous text-only baseline with the addition of vision, PDF input, tool use, JSON modes, reasoning capabilities, and prompt caching. These represent substantial functional enhancements to the model's utility across diverse workflows. The core academic performance established in the baseline appears maintained, though no new benchmark scores are available for this window to confirm performance trends. The additions of structured output formats through json_mode and json_schema address common integration needs, while tool support enables agentic workflows that were previously unavailable. Vision and PDF input capabilities extend the model's applicability to multimodal tasks. Prompt caching should improve efficiency for repetitive workflows with shared context. Users gain a notably more versatile model compared to the baseline, though the absence of updated performance metrics means stability of core capabilities cannot be verified. The breadth of new features positions this release as a major capability expansion rather than an incremental refinement.