Overview / Description
Claude Opus 4.8 is an AI large language model from Anthropic, released May 28, 2026, aimed at developers, agentic coding workflows, and complex reasoning tasks. It keeps the same pricing as Opus 4.7 while improving across benchmarks. A headline feature is Effort Control, which lets you tune how much computational effort the model spends, trading faster, cheaper responses for higher-quality ones. In Claude Code it adds Dynamic Workflows, where the model plans work and runs hundreds of parallel subagents in one session for large codebase migrations. The Messages API now accepts Mid-Task Instructions, letting you insert system entries inside the messages array to update instructions without breaking the prompt cache. Anthropic reports it is about four times less likely than its predecessor to let code flaws pass unremarked, and it is the first model to exceed 10% on all-pass legal benchmarks. As an AI model API, Claude Opus 4.8 suits coding agents, legal analysis, and computer-use and browser-agent tasks. Standard pricing is $5 per million input tokens and $25 per million output tokens.
Used For
Developers use Claude Opus 4.8 for agentic coding, codebase migrations, legal analysis, and computer-use tasks via the API and Claude Code.
Pricing
Pros & Cons
Pros
- Effort Control to trade speed and cost against output quality
- Dynamic Workflows in Claude Code run hundreds of parallel subagents per session
- Mid-Task Instructions update system guidance without breaking prompt cache
- First model to exceed 10% on all-pass legal benchmarks; ~4x less likely to pass code flaws unremarked
Cons
- Premium per-token pricing for high-volume use
- Fast Mode costs double standard rates ($10/M in, $50/M out)
- Best features require the Claude Code or Messages API ecosystem
Questions & Answers
Alternatives
OpenAI GPT-5, Google Gemini 2.5 Pro, Claude Opus 4.7, xAI Grok, DeepSeek