How do you distill a closed model? Reddit probes Anthropic's claims against Chinese rivals
LaughLegit7275 · reddit · 2026-09-19
A Reddit user asks how vendors like Anthropic and OpenAI, whose models were never open-sourced, can be "distilled" by competitors — a charge Anthropic has repeatedly leveled at Chinese labs. The post contrasts open-source distillation with the commonly assumed route for closed models: mass-generating training pairs via APIs, a practice vendor ToS explicitly prohibit, and questions whether such usage amounts to "stealing."
More from Models
- Want US frontier lab secrets? Just look at Chinese SOTA models, quips AI insider — gowthami_s · 2026-09-19
- US products quietly build on Chinese open-weight models as one firm cuts spend by ~100x — generativist · 2026-09-19
- Eval model Jev goes free on Vercel AI Gateway until Sept 25 — cramforce · 2026-09-19
- OpenAI confirms desktop Voice mode bug when starting new sessions, fix underway — juberti · 2026-09-19
- 16-model calibration test: open-weight models almost never admit uncertainty — AlexKim · 2026-09-19
- Mystery Model Jev Launches Claiming 200x Speed and 400x Cost Cuts, Devs Impressed — multiply_matrix · 2026-09-19