Amazing Fully open model of the Kimi-Linear/Qwen 3.5 generation, as far as architecture is concerned. I think Olmo is shaping up to be a pillar of LLM research on par with DeepSeek and Qwen, in its own complementary way.