3 stories
Ubicloud measured 92% schema accuracy on Kimi K2.6 tool calls with vLLM 0.28, about 2 failures in 25. Turning on constrained decoding explicitly brought it to 100% on the K2VV ToolCall benchmark.
The thread's pushback: the fix was a configuration, not a patch.
The top HN threads: facts from the discussions, both sides, insider voices. Email, Telegram, or RSS.
The release covers one model, DeepSeek-V4.1-Flash, and runs on NVIDIA SM100 GPUs and Huawei Ascend 950 NPUs. It deliberately leaves out the KV cache system and prefill/decode disaggregation.
In the thread, one reader wants to know if the whole release is a bet against US valuations. Another answers that investors don't believe Western firms will run Chinese models.
Tobi Knaup argues that capable open-weight models are becoming platforms rather than merely self-hosted alternatives. The emerging stack already includes vLLM, SGLang, llama.cpp, Ollama, MLX, fine-tunes, adapters, and deployment tooling.
HN readers liked the interoperability argument but challenged the analogy’s missing ingredient: frontier model training demands capital on a scale that open-source software never did.
