Mercury 2, a new diffusion-based language model, demonstrates superior performance compared to StepFun's 3.5 Flash model on OpenClaw benchmark tasks.

Mercury 2, built using diffusion architecture, outperforms StepFun 3.5 Flash on OpenClaw evaluation tasks

Performance comparison data is visualized through PinchBench's radar graph interface for model evaluation

The benchmark results suggest diffusion-based LLMs may offer competitive advantages over traditional transformer-based approaches