Anthropic launches Claude Opus 4.8 with better coding and reasoning

Anthropic just released their new flagship model, Claude Opus 4.8, focusing on better coding, agent tasks, and reasoning. It's a small update over Opus 4.7, with the same price. Early testers say it's more reliable and sharper, actively asking questions and spotting its own mistakes in complex tasks.

Official tests show Opus 4.8 is four times less likely to leave code bugs unmentioned, and it's better at flagging uncertainty. It scores higher on supporting user autonomy and lower on deceptive behavior. A new 'effort' control on claude.ai lets users choose between quality and speed.

In benchmarks, Opus 4.8 scored 69.2% on SWE-Bench Pro, beating GPT-5.5 and Gemini 3.1 Pro, though GPT-5.5 still leads in terminal coding. Fast mode runs 2.5x faster and costs one-third of before.