Claude Opus 4.8
Claude Opus 4.8 followed Opus 4.7 at the same standard token prices. It remains an active earlier model for code, document and agent workflows.
Updating instructions during an agent task
The launch added system entries inside the Messages API message array. A harness can use them to update permissions or environment context during a task while retaining the prompt cache. The separate dynamic-workflows preview belonged to Claude Code; it was not an automatic behavior of a single model request.Introducing Claude Opus 4.8
Specifications & access
| Detail | Value and conditions |
|---|---|
| Launch date | |
| Access | API |
| Model identifiers | claude-opus-4-8 |
| Context window | 1,000,000 tokens Direct Claude API context limit Claude Opus 4.8 model reference |
| Maximum output | 128,000 tokens Synchronous Messages API maximum Claude Opus 4.8 model reference |
| Maximum output | 300,000 tokens Message Batches only, with output-300k-2026-03-24 beta header Claude Opus 4.8 model reference |
| Input price | 5 USD per million tokens Standard direct API input; cache, batch and tool charges are separate Claude Opus 4.8 model reference |
| Output price | 25 USD per million tokens Standard direct API output Claude Opus 4.8 model reference |
| Input and output | Text and images in; text outClaude Opus 4.8 model reference |
| Reasoning control | Adaptive thinking; default effort is high.Claude Opus 4.8 model reference |
| Reliable knowledge cutoff | January 2026Claude Opus 4.8 model reference |
Related models
- Claude Opus 5
Anthropic · Claude Opus
A July 2026 Opus model for reasoning, coding and visual understanding.
Launch date API - Claude Opus 4.7
Anthropic · Claude Opus
Claude Opus 4.7 is an active coding and reasoning model with higher-resolution image processing than Opus 4.6.
Launch date API