Claude Opus 4.8 Launches: Dynamic Workflows Outperform GPT-5.5 in Key Benchmarks

Just 41 days after releasing Opus 4.7, Anthropic has officially launched Claude Opus 4.8. The new model outperforms GPT-5.5 by over 10 points on the SWE-Bench Pro benchmark and ranks first on the Artificial Analysis Intelligence Index with a score of 61.4, ahead of GPT-5.5 at 60.2. This release is considered a genuine architectural upgrade rather than a simple model iteration.

The standout feature of Opus 4.8 is the introduction of Dynamic Workflows, a tool that enables Claude to plan large tasks and distribute work across dozens to hundreds of parallel subagents, then verify outputs and return complete results. The model also achieves a 4x improvement in honesty, meaning it more accurately communicates its own uncertainty to users. In long-context tasks, Opus 4.8 significantly outperforms both GPT-5.5 and Gemini 3.1 Pro.

The new model is now available via API, claude.ai, and Claude Code, with pricing unchanged from Opus 4.7: $5 per million input tokens and $25 per million output tokens. Anthropic has also tripled the price reduction for the Fast Mode offered to team users, aiming to boost competitiveness in latency-sensitive scenarios. Industry observers believe Dynamic Workflows could redefine the operating paradigm for AI coding tools.

Editorial Note: The rapid iteration and architectural innovation behind Opus 4.8 suggest Anthropic is accelerating its technological lead over competitors. Dynamic Workflows shift the paradigm from "single-model intelligence" to "multi-agent collaboration," potentially becoming key infrastructure for the next generation of AI applications.