[ilia-a]

That doesn't seem right and seems to miss GLM 5.1 and Kimi 2.6. Not to mention there is the whole argument of cost/value that Chinese OSS models have vs GPT/Claude.