AI model
DeepSeek's V4 family of open-weight large language models, including Flash and V4.1 variants, with million-token context.
Salvatore Sanfilippo's open-source ds4 engine runs a short list of large open-weight models locally by compressing their routed experts to about two bits. Even compressed, the supported builds need high-memory Macs or GPU systems that most people do not already own.
Reality
PyTorch says vLLM's frontier models now ship as hardware-specific flat definitions that torch.compile cannot trace, and the new HW agnostic layers are what users on other accelerators get instead. The overhead figure came from an H100.
Reality
Red Hat clocks the same 20-call agent task at roughly 45 seconds on a slow backend and about 13 on a fast one. Model choice for agents is turning into a per-call latency budget, with capability as one input.
Reality
A custom SGLang framework feeds teacher outputs into the loss while training runs. The reported eightfold speedup comes from a stack of separate optimizations, and the named multipliers reach about 5.5x on their own.
Reality
ByteDance's Seed team and collaborators had 18 models write their own agent scaffolding and then rewrite it from task feedback. The held-out gains they report land inside the fluctuation band of their own evaluation.
Reality
A White House memo calls Chinese model distillation deliberate and industrial in scale, then answers it with information sharing, best practices and a demarche. The detection work stays with providers.
Reality
OpenRouter says DeepSeek went from 9% to 18% of tokens in six months, with agent loops doing most of the work. Token share and spend share are not the same league table.
Reality
SemiAnalysis's AgentX replays recorded coding-agent sessions instead of fixed prompts. On that traffic, the reported Nvidia-AMD cost gap grows as the interactivity target rises.
Reality