Anthropic released Claude Opus 4.8 in July 2026 as its most capable model for enterprise agentic work and complex coding tasks. The upgrade builds on Opus 4.7 with performance improvements and new enterprise-focused features, including dynamic workflows and richer admin controls.

What’s new in Opus 4.8

Dynamic workflows launched in research preview for Claude Code users on Enterprise, Team, and Max plans. The feature lets Claude plan work, run hundreds of parallel subagents in a single session, verify outputs, and report back. For enterprise teams, this means scaling coding and research tasks across multiple AI agents running simultaneously.

The model supports a 1M token context window by default across Claude API, Amazon Bedrock, Google Cloud, and Microsoft Foundry. It can output up to 128k tokens, supporting long-form reasoning and output without truncation. The minimum cacheable prompt dropped from 2,048 tokens to 1,024 tokens, letting smaller prompts benefit from cache savings.

Enterprise admin features

Claude Enterprise now includes richer analytics dashboards, model-level access controls, and spend alerts. Admins can see which teams use which models, track costs by department, and set guardrails to prevent budget overruns. That kind of visibility matters for large organizations managing multiple AI projects.

Pricing stayed flat from Opus 4.7: $5 per million input tokens, $25 per million output tokens. Fast mode pricing is $10 input and $50 output. No surprises on the bill side.

Why enterprises care

Claude Opus 4.8 signals that reasoning models are becoming production infrastructure, not experiments. Multi-agent workflows, token caching, and admin controls all point to the same shift: AI isn’t a feature bolt-on anymore—it’s how teams build.

When an AI company adds admin dashboards and spend alerts, enterprises start writing bigger checks.