AI TRENDS | Huawei and Partners Launch Claw-Anything AI Benchmark with GPT-5.5 Scoring 34.5%

Huawei, in collaboration with three partner institutions, has unveiled Claw-Anything, a new personal-assistant AI benchmark. According to NS3.AI, the benchmark evaluates AI models, with GPT-5.5 achieving a pass@1 score of 34.5%. Additionally, the team introduced a data pipeline that generated 2,000 training environments. This development led to a 23.7% improvement in task success for Qwen3.5-27B following fine-tuning.