Inkling Reportedly Distilled From Kimi; Distillation Called a Powerful Initialization Trick
teortaxesTex · x · 2026-11-21
teortaxesTex traces the lineage of TML's Inkling: the final model was distilled from a larger Inkling trained with heavy compute plus RL on 100K+ sequences, and the original Inkling was distilled from Kimi. His takeaway: distillation is a very powerful trick for initializing a new project — bootstrap from an existing frontier model, then push further with your own compute and RL.
More from Models
- Gemini 4 'Cobalt' leak debunked: internal codename for already-released Gemini 4 Neon — lyraxana · 2026-11-21
- "The 'it' in AI models is the dataset" resurfaces, fueling frontier lab data debate — teortaxesTex · 2026-11-21
- Dev runs Gemini coding all day without hitting Pro limits; agent nails sound design — lipeng0820 · 2026-11-21
- Google appears to reset Ultra weekly limits for all Antigravity users — novice_invader · 2026-11-21
- Claims surface: Navier-Stokes and Riemann proofs via LEAN, by OpenAI and Claude — BasketVolleyball11 · 2026-11-21
- Yacine: 'It might be over for OpenAI — and the killer isn't Claude, it's Gemini' — yacineMTB · 2026-11-21