Chinese AI company DeepSeek has released version 3.1 of its flagship large language model, expanding the context window to 128,000 tokens and increasing the parameter count to 685 billion. The update ...