Krafton open-sources Raon-1 nano, a 1.5B on-device LLM distilled from MoE
DimitrisPapail · x · 2026-11-11
KRAFTON released Raon-1 nano under Apache 2.0 on Hugging Face: a dense LLM with 1.5B effective / 3.1B total parameters, built by pruning and distilling its 30B-A4B Raon-1 mini preview via a MoE-to-Dense approach (NeurIPS 2026 Oral), then trained on 3T tokens plus SFT, OPD, and RL. Designed for Korean/English conversation, instruction following, and tool use on device, continuing its PUBG Ally → on-device LLM journey. Published benchmarks show it competitive with or ahead of Qwen3.5 2B, Gemma 4 E2B, MiniCPM5 2B, and LFM3 1.7B on IFBench, MultiChallenge, AndroidControl, and Korean chat tasks.
More from Models
- Gemini 4 Carbon matches predecessor's score but explores less, gets distracted — patience_cave · 2026-11-11
- Fable 5.2 takes MazeBench frontier, matches GPT-6 Astra puzzle-solving — patience_cave · 2026-11-11
- MazeBench: Fable 5.2 scores 15%, near-frontier at half price; Gemini 4 Carbon 11% — patience_cave · 2026-11-11
- ERNIE 5.5 Flash Preview enters LMArena Battle Mode for crowdsourced ranking — arena · 2026-11-11
- Blind tests: DeepSeek matches Opus on 95% of agentic loops at 40x lower cost — bindureddy · 2026-11-11
- Indie dev flips from $20 Pro back to $200 Gemini Ultra after two days of deep use — yihui_indie · 2026-11-11