GPT-5.6 Sol outperforms Claude Mythos 5 in coding benchmarks. Despite the technical win, the US government is forcing a restricted rollout of the model. OpenAI labels these access rules unsustainable. This friction between state oversight and deployment speed creates a precarious bottleneck for developers awaiting full model capabilities.