Overview
- Anthropic launched Claude Opus 4.8 with effort controls that let users set how much compute the model uses, dynamic workflows for Claude Code that can spawn parallel sub‑agents, and live update capabilities in the Messages API so developers can change instructions and limits without restarting agents.
- The company says Opus 4.8 improves coding, reasoning, and agentic performance while adding safety measures to lower deceptive behavior and reduce compliance with misuse attempts.
- Anthropic kept token pricing the same as the prior release, with standard rates of $5 per million input tokens and $25 per million output tokens and faster mode at higher rates, which leaves enterprise token bills a central operational issue as usage scales.
- Fast Company reported that Anthropic has been rapidly scaling capital and infrastructure, raising $65 billion at about a $965 billion valuation, securing large external compute capacity including a Colossus arrangement with SpaceX/xAI reported at roughly $1.25 billion per month, and filing a confidential draft S‑1 with the SEC as it prepares for an IPO.
- Claude Code’s earlier jump in capability with Opus 4.5 helped drive major enterprise adoption and rapid reported ARR growth, so Opus 4.8 is positioned to deepen automation for large customers even as it increases exposure to rising compute costs and concentrated supplier risk.