Anthropic introduced Claude Opus 4.8 on May 28, 2026 as a direct upgrade to Opus 4.7, carrying forward the same Opus lineage while emphasizing benchmark gains and what the company describes as "a more effective collaborator." Early access testers highlighted noticeably better judgment when the model is performing agentic work, particularly inside Claude Code, where it is being used to tackle very large-scale problems through a new dynamic workflows capability. Anthropic simultaneously released an Opus 4.8 System Card that lays out a broader capability evaluation profile spanning coding, agentic skills, reasoning, and practical knowledge tasks, giving practitioners a reference point for how the model compares to its predecessor and peers.

The release also reshaped how teams can work with Opus at speed: a new user-facing effort control on claude.ai lets people dial up or down how much the model invests in a response, while fast mode now runs Opus 4.8 at roughly 2.5× normal speed at one-third the prior per-token cost, making latency-sensitive workflows more economical. Anthropic's own platform documentation frames Opus 4.8 as the baseline from which the newer Opus 5 draws its largest gains in deep reasoning, agentic long-horizon tasks, and test-time compute scaling, which underlines Opus 4.8's role as a stable, broadly capable workhorse well suited to complex coding sessions, structured analysis, and tool-driven research pipelines where reliable judgment matters more than headline novelty.