KRAFTON distills 29B MoE into 1.5B Raon-1 Nano to power game NPCs
Sam Witteveen · youtube · 2026-11-17
Sam Witteveen reviews KRAFTON's Raon-1 Nano, a 1.5B dense LLM distilled from a 29B MoE specifically to power game NPCs. Key points:
- Parent Raon-1 is an MoE model, Apache 2.0 licensed, with a free API and local deployment
- Covers benchmarks, the MoE-to-dense distillation process (drawing parallels to Karpathy's nanochat), and a live NPC demo
- Discusses why a game company like the PUBG maker wants its own in-house models
More from coding & agent
- Developer sees little value in extreme multi-agent coding workflows so far — DanielLockyer · 2026-11-18
- Project triangle flips in the AI agent era: defining 'good' becomes the job — lak_luster · 2026-11-18
- Every's vibe check: Gemini 4 Carbon codes better but judgment lags Fable 5.2 — every · 2026-11-18
- SWE-CoWork benchmark: best model fully passes only 8 of 101 handoff coding tasks — omarsar0 · 2026-11-18
- LangChain explains Agent vs. Deep Agent vs. Managed Deep Agent — LangChain · 2026-11-18
- Dev doubts extreme multi-agent workflows: little value, multiple subscriptions burned — DanielLockyer · 2026-11-17