whatmodel.sh / Raon-1 nano

Raon-1 nano is an open-weights language model from KRAFTON, released November 2026. Its published size is 3.1B parameters (1.5B active per token) and a 262k-token context window. Licence: apache-2.0.

Catalogue identity notes

KRAFTON AI's Korean/English dense model, derived from Raon-1 mini preview (30B-A4B) by MoE-to-Dense pruning and distillation, then SFT, on-policy distillation and RL; thinking mode by default. Repo created 2026-10-29, Apache 2.0 declared 2026-11-09 (commit f3006ef), current weights committed 2026-11-10 (commit 0bed13e 'Update Raon-1 nano model weights'); benchmark table added after that weight update and revised 2026-11-11 (pinned sha fcf2d1d). released set to 2026-11 because the licence and the current weights date from November; earlier October weights (RL step 440, OPD step 169) are not this checkpoint. Results are KRAFTON's own column; baselines are other labs' models and not used.

Model information sources

Release date ↗ · Weights licence ↗ · Parameter count ↗

Dense model. Card: 3.1B total including a 1.5B per-layer embedding table, 1.5B effective (active_b follows the Gemma E-model convention of recording the effective count). Checkpoint also holds a 55M MTP module (3.2B). The GGUF card at 4534f1f still prints the older wording '3.2B parameters, including a 1.6B per-layer embedding table'.

No comparable score yet. Its published results are below, but they don't overlap enough with other models to calculate a score.

$ how it compares

Scored LLMs released closest to Raon-1 nano, out of 1206 with a score.

#modelscorewhereRAMreleased
1Claude Fable 5.2 Anthropic89±1api $16.00/M2026-11
14Gemini 4 Hydrogen Google79±2api $3.00/M2026-11
22Hy4 Tencent78±1api price unknown2026-11
32Muse Ember Meta76±1api $0.20/M2026-11
37Inkling 2 Small (594B-A17B) Thinking Machines75±1api price unknown2026-11

see all LLMs · find the ones that fit your machine

$ what it's good at

other tests4 not in a task · 0.0% of its score
testresultvs the others in its rundifficultycountswhy
IFBench (aggregation unspecified) lab75.9% ✎▂▆▂▃▄▄▂▃▄▆█▇ beats 69/88—not counted yet: no other model in the same run to compare it with
MMLU-Redux lab81.1% ✎▁▁▃▁▂▄▄▅▇▅█▄ beats 14/64—not counted yet: no other model in the same run to compare it with
MultiChallenge lab40.5% ✎▁·▁·▁▅█▃▅▃█▃ beats 5/33—not counted yet: no other model in the same run to compare it with
FACTS Grounding lab44.4% ✎▄▄·····▄·█·▄ beats 1/5—not counted yet: no other model in the same run to compare it with

$ where to run it

Memory is what a version takes in GPU memory (or a Mac's unified memory) with an 8k-token conversation. With less, it runs partly from system RAM, more slowly. Enter your machine to see what runs at full speed.

Other ways to run it

version

  1. Run it as published · needs 3.7 GB memory, 2.08 GB download · from huggingface.co/KRAFTON (official)
    git clone -b raon-1-nano-preview https://github.com/krafton-ai/llama.cpp && cd llama.cpp && cmake -B build && cmake --build build --config Release -j --target llama-cli llama-server && ./build/bin/llama-server -hf KRAFTON/Raon-1-nano-gguf:Q4_K_M --ctx-size 65536 --reasoning on --chat-template-kwargs '{"enable_thinking": true}'

Model repository and documentation: huggingface.co/KRAFTON ↗

$ published results · 4 results, not enough overlap for a score
testresultsource and conditions
MMLU-Redux81.1%KRAFTON/Raon-1-nano model card · maker-reported

KRAFTON own evaluation, pinned Raon-1-nano card, 'Benchmark Results' HTML table, column 'Raon-1 nano (E1.5B)'. Card footnote: 'Raon-1 nano sampling: temperature 0.7 for Knowledge, temperature 1.0 for Tool Use, temperature 0.1 with presence penalty 1.5 for the others.' Thinking mode is the card's default (mode not labelled per row). Row 'Knowledge / MMLU-Redux (temperature 0.7)', printed 81.1.

IFBench (aggregation unspecified)75.9%KRAFTON/Raon-1-nano model card · maker-reported

KRAFTON own evaluation, pinned Raon-1-nano card, 'Benchmark Results' HTML table, column 'Raon-1 nano (E1.5B)'. Card footnote: 'Raon-1 nano sampling: temperature 0.7 for Knowledge, temperature 1.0 for Tool Use, temperature 0.1 with presence penalty 1.5 for the others.' Thinking mode is the card's default (mode not labelled per row). Row 'Instruction Following / IFBench (aggregation not stated; temperature 0.1, presence penalty 1.5)', printed 75.9.

MultiChallenge40.5%KRAFTON/Raon-1-nano model card · maker-reported

KRAFTON own evaluation, pinned Raon-1-nano card, 'Benchmark Results' HTML table, column 'Raon-1 nano (E1.5B)'. Card footnote: 'Raon-1 nano sampling: temperature 0.7 for Knowledge, temperature 1.0 for Tool Use, temperature 0.1 with presence penalty 1.5 for the others.' Thinking mode is the card's default (mode not labelled per row). Row 'Instruction Following / MultiChallenge (temperature 0.1, presence penalty 1.5)', printed 40.5.

FACTS Grounding44.4%KRAFTON/Raon-1-nano model card · maker-reported

KRAFTON own evaluation, pinned Raon-1-nano card, 'Benchmark Results' HTML table, column 'Raon-1 nano (E1.5B)'. Card footnote: 'Raon-1 nano sampling: temperature 0.7 for Knowledge, temperature 1.0 for Tool Use, temperature 0.1 with presence penalty 1.5 for the others.' Thinking mode is the card's default (mode not labelled per row). Row 'Context Following / FACTS Grounding (temperature 0.1, presence penalty 1.5)', printed 44.4.

$ improve this page

Use a short GitHub form or edit the data through a pull request. Every number needs a source; here's which ones we can use.

data updated · sources and licences · json