Claude Opus 4.8: Fewer Code Flaws, Faster Performance

Anthropic has launched Claude Opus 4.8, a new AI model that is four times less likely to pass code flaws unnoticed compared to its previous version. This update comes just six weeks after Opus 4.7. Opus 4.8 scores 69.2% on SWE-Bench Pro, outperforming other models like GPT-5.5 and Gemini 3.1 Pro. The model shows improvements in areas like agentic coding, multidisciplinary reasoning, and financial analysis. Knowledge work scores also increased from 1753 to 1890. Anthropic says Opus 4.8 has "sharper judgement" and can work independently for longer. Early testers noted it flags uncertainties more often and makes fewer unsupported claims. The model also shows gains in prosocial traits, with deception rates lower than Opus 4.7. Pricing remains the same. A new "Fast mode" now runs 2.5 times faster and costs three times less. Three new features are also included. Dynamic Workflows allows Claude to plan work and run many parallel subagents. Effort Control gives users a slider to adjust processing power for responses. The Messages API now accepts system entries mid-task. This rapid development shows the ongoing advancements in AI capabilities, which could impact how we interact with technology.