Anthropic Releases Claude Opus 4.8 with Enhanced Capabilities and Lower Pricing

Anthropic has released Claude Opus 4.8, an upgraded version of its Claude Opus model that offers improvements across multiple performance benchmarks and serves as a more effective collaborative tool. The new model is available immediately at the same price as the previous version, making it a direct upgrade for existing users.

The release includes several new features designed to enhance user experience and capability. Users on claude.ai now have granular control over the computational effort Claude allocates to tasks. Claude Code introduces a "dynamic workflows" feature enabling the model to handle very large-scale problems, while a new fast mode for Opus 4.8 operates at 2.5 times the standard speed at one-third the cost of fast mode pricing for previous models.

Clause Opus 4.8 demonstrates significant improvements in agentic capabilities, coding, reasoning, and practical knowledge work tasks compared to Opus 4.7. Early testers report that the model exhibits better judgment, improved error detection, and more reliable task completion. Notable performance gains include achieving 84% on Online-Mind2Web (a computer use and browser agent benchmark), being the first model to exceed 10% on the Legal Agent Benchmark's all-pass standard, and completing every case end-to-end on the Super-Agent benchmark—outperforming prior Opus versions and GPT-5.5 at cost parity.

Practical improvements emphasized by early users include more efficient tool calling that requires fewer steps, stronger context retention across extended sessions, enhanced consistency in following instructions for autonomous engineering workloads, and proactive identification of input and output issues. The model shows meaningful improvements in consistency and reasoning quality for professional workflows, with reviewers noting noticeably higher quality analysis, faster execution times, and improved information density compared to predecessor models.