Kimi K3.1 Pro saves 55% vs Opus 5.5 but trails on precision

A 13-task long-horizon test found Moonshot's Kimi K3.1 Pro cuts costs by an average of 55% versus Opus 5.5, but falls short on accuracy and task completion. Kimi leads in creative and visual work, while Opus wins on computation, code-based 3D modeling and front-end development.

• Kimi K3.1 Pro averages 55% cost savings, below its advertised halving of Opus pricing
• Kimi excels in Blender scenes and creative visuals; Opus wins on numerical precision
• Higher API message counts and cached-token rereads erode Kimi's savings
• In code-only 3D modeling Kimi often stops early and returns incomplete output