ITdaily reported on 29 May 2026 that Anthropic had introduced Claude Opus 4.8. This is trade press rather than Anthropic's own page, so confirm details with the vendor before relying on them. We use Claude ourselves in some work and have no wish to overstate a release; the points below are limited to what the article states.
Points reported
- Pricing is the same as for Opus 4.7: $5 per million input tokens and $25 per million output tokens.
- A 'fast mode' is priced at one-third of previous models' rates.
- Claude Code gains 'dynamic workflows' that split large tasks into many parallel sub-agents which verify results, available on Enterprise, Team and Max plans.
- Users can adjust reasoning effort, with high as the default.
What it means for buyers
- Flat list prices for a stronger model make re-testing worthwhile. Re-run your own evaluation set instead of assuming the improvement applies to your task.
- Parallel sub-agents shift cost and risk: more work happens without a person watching, so budgets, logs and review gates matter more.
- Keep model names in configuration, so you can switch without code changes.
The article gives no benchmark scores or availability details. Measure on your own data.