Anthropic has released Claude Opus 4.8, an upgrade to Opus 4.7 that ships at the same base price. For builders running agentic workloads, two numbers stand out immediately: fast mode now runs at 2.5x the speed, and that fast mode costs three times less than it did for previous Opus models.
The model improves across coding, agentic skills, reasoning, and practical knowledge work. 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 cost parity, according to early tester Kay Zhu. On Online-Mind2Web, it scores 84%, a meaningful jump over both Opus 4.7 and GPT-5. It is also the first model to break 10% overall on the all-pass standard of a Legal Agent Benchmark, which testers describe as a direct signal for how much real attorney work can be handed off with confidence.
On CursorBench, Opus 4.8 exceeds prior Opus models across every effort level. Tool calling is more efficient, using fewer steps for the same intelligence. That matters a lot when you are paying per step in a long-horizon coding agent.
Three new product features ship alongside the model. Users on claude.ai can now control how much effort Claude puts into a task. Claude Code gains a "dynamic workflows" feature designed to tackle very large-scale problems. And fast mode, already mentioned on price, is now the default path to higher throughput for teams that need it.
Early testers from Cursor, Harvey, and other teams emphasize that the quality jump is mostly about judgment. The model catches its own mistakes, pushes back on unsound plans, and builds confidence before making large changes in multi-service environments. For long sessions where voice, style, and technical execution all have to stay consistent, testers report Opus 4.8 holds context better than 4.7 did.
Full benchmark details are in the Claude Opus 4.8 System Card linked from the announcement.
What to do today: If you are running Claude Code or a browser-agent pipeline, swap in Opus 4.8 and enable fast mode. The cost reduction on fast mode alone makes it worth re-benchmarking your existing workflows. If you use Claude Code for large refactors, test the new dynamic workflows feature on a real large-scale task to see whether it reduces the manual checkpointing you are likely doing today.