# Krafton releases Raon-1 mini preview, a 30B-A4B MoE model with 262,144 token context

Digest AI · Generative AI & Models · published 2026-10-21T07:54:08Z

Canonical: https://digestai.news/story/krafton-releases-raon-1-mini-preview-a-30b-a4b-moe-model-with-262-144

## Summary

KRAFTON AI announced a preview of its Raon-1 mini language model, a 30B-A4B Mixture‑of‑Experts system built from scratch. The model uses 18 linear attention layers and 7 full attention layers, and it supports up to 262,144 tokens of context. It runs on a single NVIDIA RTX 5090 GPU (32 GB VRAM) and offers quantization in MXFP8 and NVFP4 formats.

Raon-1 mini’s sparse MoE activates six of 128 routed experts plus one shared expert per token, using 4.3B of the 28.8B total parameters. The preview includes a DFlash2 draft model for speculative decoding, delivering up to 2.18× higher output throughput on a single NVIDIA B300 compared with autoregressive decoding. The model is released under the Apache 2.0 license, with open‑weight weights, and can be accessed via a free limited‑time playground or through hosted API access after agreeing to separate Hosted Access Terms. The release is labeled a preview, and the article notes that specs and results may change before the final version.

## Key points

- Raon-1 mini is a 30B-A4B MoE model with 18 linear and 7 full attention layers
- Supports up to 262,144 tokens of context and runs on a single NVIDIA RTX 5090 (32 GB VRAM)
- Sparse MoE activates 6 of 128 experts plus 1 shared, using 4.3B of the 28.8B parameters per token

## Why it matters

A long‑context, multilingual open‑weight model gives developers a free alternative for Korean‑centric AI applications and large‑scale reasoning.

## Sources

1. [KRAFTON/Raon-1-mini-preview · Hugging Face](https://huggingface.co/KRAFTON/Raon-1-mini-preview) (huggingface.co, 2026-10-21, primary source)

## Cite

Digest AI, "Krafton releases Raon-1 mini preview, a 30B-A4B MoE model with 262,144 token context", 21 October 2026, https://digestai.news/story/krafton-releases-raon-1-mini-preview-a-30b-a4b-moe-model-with-262-144

---

Written by Digest AI's editorial model from the linked sources; the sources are the record. Headlines, digests and key points are written by Digest AI and may be quoted with a link to the story page. Linked articles belong to their publishers. Terms: https://digestai.news/terms#reuse
JSON: https://digestai.news/story/krafton-releases-raon-1-mini-preview-a-30b-a4b-moe-model-with-262-144.json