New benchmarks show AI rapidly gaining scientific research intuition while still struggling with visual reasoning. Western open-weight models from Reflection and Mistral offer higher efficiency but trail Chinese competitors, as the US government experiments with distributed supercomputer training.