Anthropic Releases Claude Opus 4.8 with Major Gains in Reasoning, Agents, and Legal Work
Key Takeaways
- ▸Claude Opus 4.8 delivers across-the-board performance improvements with the highest-ever Legal Agent Benchmark score and superior agentic capabilities, positioning it as the strongest model for enterprise and professional workflows
- ▸Three accompanying features lower barriers to adoption: effort control on claude.ai, dynamic workflows in Claude Code for complex problems, and 3× cheaper fast mode with 2.5× speed
- ▸Available immediately at the same price as Opus 4.7, making performance gains a direct upgrade for existing users without cost increases
Summary
Anthropic has unveiled Claude Opus 4.8, a significant upgrade to its flagship large language model launching today at the same price as Opus 4.7. The new version delivers across-the-board improvements in coding, reasoning, agentic capabilities, and practical knowledge work, with early testers praising its enhanced judgment and reliability for complex, multi-service problems.
Three new features accompany the launch: Claude.ai users gain granular control over effort levels, Claude Code introduces "dynamic workflows" for large-scale problem-solving, and fast mode for Opus 4.8—delivering 2.5× speed—is now three times cheaper than previous iterations. The model achieves the highest-ever score on the Legal Agent Benchmark, breaking the 10% all-pass threshold for the first time, while also outperforming competitors on agent benchmarks and browser automation tasks.
Benchmark results show Opus 4.8 as the only model to complete every case end-to-end on Anthropic's Super-Agent benchmark, beating GPT-5.5 at parity on cost. It achieves 84% on Online-Mind2Web (browser/computer use), meaningfully ahead of prior versions. Early adopters across legal, engineering, and autonomous systems report tangible improvements in reliability, consistency, and the model's ability to proactively flag issues.
- Early testers across legal, engineering, and autonomous systems report significantly improved reliability, judgment, and context management—critical factors for high-stakes professional and agentic workloads
Editorial Opinion
Claude Opus 4.8 represents a calculated narrowing of the capability gap with competitors, with Anthropic showing particular strength in domains that demand reliability: legal work, autonomous agents, and long-context reasoning. The emphasis on judgment and error detection—rather than raw speed—signals a maturation toward enterprise trust, which matters more in professional services than raw benchmarks. The 3× price cut on fast mode is smart positioning: it defends against commoditization at the speed tier while preserving Opus 4.8's flagship status.
