Overview
Claude Opus 4.8 is Anthropic's flagship as of late May 2026, building on Opus 4.7 with improvements across benchmarks at the same price — $5 per million input tokens and $25 per million output tokens. It's the model teams escalate to when correctness and patience matter more than cost, especially inside Claude Code for long autonomous engineering runs.
The headline additions are dynamic workflows in Claude Code (multi-step, self-directed sessions that run hundreds of parallel subagents on codebase-scale migrations) and an effort-control feature on claude.ai that lets you dial reasoning depth up or down per task. A new fast mode runs roughly 2.5× faster at about a third of the previous cost, making the Opus tier more practical for interactive work. Opus 4.8 is also notably more honest — roughly four times less likely than its predecessor to let flaws in code it has written pass unremarked.
Key capabilities
- ~4× less likely to pass code flaws unremarked
- Dynamic workflows with hundreds of parallel subagents
- Effort control on claude.ai; 2.5× faster, 3× cheaper fast mode
- 84% Online-Mind2Web (computer-use and browser agent)
At a glance
Online-Mind2Web
84%
Honesty
~4× fewer unchecked code flaws
Context
200K tokens
Pricing: Premium Opus pricing — $5 / 1M in, $25 / 1M out (unchanged from 4.7); fast mode $10 / $50.
Pros & cons
What we like
- Best-in-class reasoning and autonomous coding
- Dynamic workflows handle long engineering sessions
- Effort control trades speed for depth on demand
Trade-offs
- Opus is still the priciest Claude tier
- Pro-level effort is latency-heavy for chat
- Overkill for routine work Sonnet handles well
The verdict
Claude Opus 4.8 is the model to escalate to when work is genuinely hard — multi-step engineering, high-stakes analysis, autonomous agents. It's a sharper, more honest collaborator with dynamic workflows for codebase-scale work; pair it with Sonnet for cost control and reserve Opus for the steps that truly need it.
Best for: Autonomous coding agents and high-stakes frontier reasoning.
Frequently asked questions
What's new in Claude Opus 4.8?
It builds on Opus 4.7 with broad benchmark gains at the same price ($5 / $25 per 1M tokens). It launches with dynamic workflows in Claude Code (hundreds of parallel subagents), effort control on claude.ai, and a fast mode that runs ~2.5× faster at roughly a third of the prior cost. Opus 4.8 is also ~4× less likely to let flaws in its own code pass unremarked.
When should I use Opus instead of Sonnet?
Use Opus 4.8 for the hardest reasoning, long autonomous agent runs and large refactors. Sonnet remains the better everyday default for chat and cost-per-quality.