[Kimi K3 Scores 69% on DeepSWE, Just One Point Behind Claude Fable 5] According to monitoring by Beating, AI data company Datacurve has added Kimi K3 to the long-term software engineering benchmark DeepSWE. This benchmark tests whether models can independently complete real, complex, long-term development tasks. Kimi K3 achieved a task pass rate of 69%. GPT-5.6 Sol and Claude Fable 5 scored 73% and 70%, respectively. Datacurve stated that Kimi K3 is the first open-weight model to reach the level of cutting-edge closed-source models on DeepSWE. DeepSWE includes 113 original tasks sourced from 91 active open-source projects, covering 5 programming languages. All models use the same set of tools. The tasks are not directly adapted from existing code submissions to minimize training data contamination. Kimi K3's average cost per task is $4.65. GPT-5.6 Sol costs $8.39 per task, while Claude Fable 5 costs $21.63 per task. [Original Link]