Claude Opus 4.8: what Anthropic's new model can do — and what it means for your business
Anthropic released Claude Opus 4.8, its most capable model yet. Beyond the benchmarks, the real question for a business is different: what automation becomes possible? Here's the short version, with official sources.
What Opus 4.8 is
Opus 4.8 is the most capable Claude model at launch (June 2026), at the frontier across three areas: coding, agentic tasks (agents that run workflows on their own), and knowledge work. It's an upgrade over Opus 4.7, with benchmark gains and — per early testers — sharper, more reliable judgment on autonomous tasks.
What it can do, in business terms
Agents that finish the job. On the Super-Agent benchmark, Opus 4.8 is the only model to complete every case end-to-end — beating prior Opus models and GPT-5.5 at the same cost. For a business: AI agents that take on multi-step workflows with fewer interruptions and human corrections.
Browser and computer automation. It's the strongest "computer use" model tested, scoring 84% on Online-Mind2Web. In practice it can operate web interfaces and apps like an operator — filling forms, extracting data, moving information between systems.
1M-token context. It can read very large documents in one pass — contracts, reports, entire knowledge bases — without chopping them up. Up to 128K output tokens means long deliverables too.
Better judgment. More reliable on agentic tasks means fewer costly mistakes when the agent works unsupervised.
What it costs
Pricing stayed at $5 / million input tokens and $25 / million output. Prompt caching saves up to 90% on repeated context, and batch processing (large, non-urgent volumes) cuts 50%. For automation at scale, the per-task cost drops significantly.
Why it matters for business
A model that's better at agentic tasks changes what you can realistically automate. Workflows that used to need step-by-step supervision — lead qualification, reporting, follow-up, support, internal operations — can be run by agents that carry the work forward on their own, with human input only at decision points. The combination of large context, strong computer use, and controlled cost makes Opus 4.8 a practical foundation for operational AI agents in a company.
Sources: ↗ Official Anthropic announcement · ↗ Claude Opus page
See what MassAI agents can do →