Anthropic just released Claude Opus 4.8, and while the benchmark improvements are quite real, the most meaningful change for everyday users is something far simpler.
You can now tell Claude how hard to think before it responds to your query. Along with that, dynamic workflows are now available in research preview for Enterprise, Team, and Max plan users.
What is effort control and why does it matter?
A new control sitting alongside the model selector on claude.ai page, where you interact with the AI model. It lets you choose the amount of effort Claude applies to any given query.
There are five effort settings to choose from when using Opus 4.8: Low, Medium, High (Default), Extra, and Max. As the name suggests, low effort responses are faster, and therefore, better for quick or trivial questions, email drafts, or anything else that requires speed over depth.
The High and Max settings, on the other hand, force Claude to think more deeply before responding. These effort levels are ideal for complex multi-step problems, detailed analysis or comparison, and anything where accuracy matters more than speed. However, as the AI model takes more time to respond, these effort levels exhaust your rate limits faster.
The control is available on all plans. Until now, you only submitted a prompt, selected the AI model based on its known strengths, and Claude decided how much thinking to apply. Effort control effectively hands that decision to you.
What else changed?
The most significant practical upgrade for developers is dynamic workflows. Claude Code can now plan a large task, spin up hundreds of parallel subagents in single session, and verify its own outputs before reporting back.
The Messages API now accepts system entries inside the messages array, allowing developers to update Claude’s instructions mid-task without breaking the prompt cache. This is especially useful for updating permissions, token budgets, or environment context.
Anthropic also mentions that Opus 4.8 is around four times less likely than its predecessor to let flaws in code pass without noticing. The release arrives 41 days after Opus 4.7, reflecting an accelerating update frequency.