Chinese AI lab DeepSeek released an updated version of its R1 reasoning model that performs well on a number of math and coding benchmarks. Some AI researchers speculate that at least a portion came ...