Anthropic Introduces Claude Opus 4.8 with Improvements for Coding and Agentic Tasks

Anthropic introduced Claude Opus 4.8, an update to its flagship model with improved judgment for agentic tasks. The model is available at the same price as Opus 4.7 and works with new features on claude.ai and Claude Code.

Three New Features with Opus 4.8

Along with the model itself, Anthropic launched tools for better control:

• On claude.ai — a slider to control effort: from fast mode to deep analysis
• Claude Code received Dynamic Workflows for solving very large tasks (taking days of work)
• Fast Mode for Opus 4.8 is now three times cheaper and runs 2.5× faster than standard

On standard benchmarks for coding, reasoning, and practical tasks, Opus 4.8 outperforms Opus 4.7 and competes with GPT-5.5.

What the Tests Showed

Opus 4.8 is unique for its reliability on sustained tasks. On the Super-Agent benchmark, it is the only model that completed all tasks end-to-end and outperformed both Opus 4.7 and GPT-5.5. On CursorBench (a code editor test), the model exceeds all difficulty levels, calling tools more efficiently. On the Legal Agent Benchmark, Opus 4.8 is the first model to cross the 10% threshold on all-pass standard, meaning lawyers can now trust it with more complex work. On Online-Mind2Web (browser agents), the model achieved 84%—a notable jump over Opus 4.7 and GPT-5.5.

What Developers Say

Engineers at Devin noted that Opus 4.8 works with tools more cleanly and follows instructions with the consistency needed for autonomous operation. The model fixed the verbosity issues that were present in 4.7.

"Opus 4.8 is a quality update: faster, easier to collaborate with, and

better maintains context and style throughout long sessions," says one early tester.

The CoCounsel team sees that Opus 4.8 provides better reasoning in complex analyses and completes work faster with denser results. On the Super-Agent benchmark, the model proved it can manage sustained workflows without errors and without human intervention.

What This Means

Opus 4.8 is not just a version update, but a signal that base models are developing very rapidly. In one quarter, Anthropic made it substantially more useful for commercial agents: Devin agents are more reliable, lawyers delegate more, engineers save time on code review. For businesses, this means investments in AI tools are becoming more profitable.