llama.cpp
UnknownChecked the exact model config and the runtime architecture registry; no matching declared architecture was found. This is unknown, not evidence of incompatibility.
Checked 2026-10-27 00:09 UTC
Checked source ↗KRAFTON · Released Oct 21, 2026 · observed 26 Oct 2026
Public-source support evidence
Raon-1-mini-preview
We have not found public evidence yet that Raon-1-mini-preview is listed in llama.cpp, Ollama, LM Studio, MLX or vLLM. We ran the Q4_K_M GGUF on one Mac; see Hardware notes below.
KRAFTON/Raon-1-mini-preview
Raon-1-mini-preview is a text generation model published by KRAFTON on Hugging Face. We have not written a summary yet; see the model card ↗.
View source model ↗SUPPORT EVIDENCE
Runtime status
Ollama
UnknownThe exact model and runtime-specific sources were checked; no positive model-level evidence was established.
Checked 2026-10-26 23:48 UTC
Version v0.40.11
Checked source ↗LM Studio
UnknownThe exact model and runtime-specific sources were checked; no positive model-level evidence was established.
Checked 2026-10-27 00:28 UTC
Checked source ↗MLX / mlx-lm
UnknownChecked the exact model config and the runtime architecture registry; no matching declared architecture was found. This is unknown, not evidence of incompatibility.
Checked 2026-10-27 00:09 UTC
Checked source ↗vLLM
UnknownChecked the exact model config and the runtime architecture registry; no matching declared architecture was found. This is unknown, not evidence of incompatibility.
Checked 2026-10-27 00:09 UTC
Checked source ↗MLX community uploads may target mlx-vlm; an upload observation does not establish mlx-lm compatibility.
MEMORY ESTIMATES
Which file should I grab?
The summary estimate of about 26 GB is the recommended Q4_K_M quant (19.7 GB observed), from file size × 1.2 + 1.5 GB. Each card is the largest observed quant whose estimate fits that capacity.
16 GB capacity estimate
Which file should I grab?
doesn't fitNo observed complete quant fits this capacity using file size × 1.2 + 1.5 GB.
24 GB capacity estimate
Which file should I grab?
IQ4_XS · 16.1 GB observed · about 21 GB ↗About 21 GB using file size × 1.2 + 1.5 GB. Fit is an estimate; compatibility, speed, context length and hardware have not been tested.
32 GB capacity estimate
Which file should I grab?
Q5_K_M · 22.2 GB observed · about 29 GB ↗About 29 GB using file size × 1.2 + 1.5 GB. Fit is an estimate; compatibility, speed, context length and hardware have not been tested.
64 GB capacity estimate
Which file should I grab?
Q8_0 · 30.7 GB observed · about 39 GB ↗About 39 GB using file size × 1.2 + 1.5 GB. Fit is an estimate; compatibility, speed, context length and hardware have not been tested.
128 GB capacity estimate
Which file should I grab?
Q8_0 · 30.7 GB observed · about 39 GB ↗About 39 GB using file size × 1.2 + 1.5 GB. Fit is an estimate; compatibility, speed, context length and hardware have not been tested.
192 GB capacity estimate
Which file should I grab?
Q8_0 · 30.7 GB observed · about 39 GB ↗About 39 GB using file size × 1.2 + 1.5 GB. Fit is an estimate; compatibility, speed, context length and hardware have not been tested.
256 GB capacity estimate
Which file should I grab?
Q8_0 · 30.7 GB observed · about 39 GB ↗About 39 GB using file size × 1.2 + 1.5 GB. Fit is an estimate; compatibility, speed, context length and hardware have not been tested.
512 GB capacity estimate
Which file should I grab?
Q8_0 · 30.7 GB observed · about 39 GB ↗About 39 GB using file size × 1.2 + 1.5 GB. Fit is an estimate; compatibility, speed, context length and hardware have not been tested.
Memory estimates are the observed file size plus fixed headroom (20% plus 1.5 GB runtime allowance); real needs also depend on context length, KV cache and runtime settings. Fit is an estimate; it does not establish runtime compatibility, speed, context length, or hardware performance. Text-weights-only estimates do not include a projector or total vision memory where unknown.
Terms: a quant is a lower-precision version of the weights; GGUF is a file format used by llama.cpp and compatible apps; MLX targets Apple silicon; vLLM is a model-serving engine for NVIDIA GPUs; the KV cache is stored attention state that grows with context length. Terms explained ↗
OBSERVED FILES
Quant uploads
Sizes and file groups come from the listed uploader repositories. A size is marked as an estimated fit only for a complete observed GGUF quant group.
| Uploader | Format & quant | Observed size | Estimated fit | Files |
|---|---|---|---|---|
| KRAFTONKRAFTON/Raon-1-mini-preview-gguf ↗Observed 27 Oct 2026 | GGUF · IQ4_XS | 16.1 GB | 24 GB, 32 GB, 64 GB, 128 GB, 192 GB, 256 GB, 512 GBEstimate only; requires the complete GGUF quant group. | 1 file |
| KRAFTONKRAFTON/Raon-1-mini-preview-gguf ↗Observed 27 Oct 2026 | GGUF · Q4_K_M | 19.7 GB | 32 GB, 64 GB, 128 GB, 192 GB, 256 GB, 512 GBEstimate only; requires the complete GGUF quant group. | 1 file |
| KRAFTONKRAFTON/Raon-1-mini-preview-gguf ↗Observed 27 Oct 2026 | GGUF · Q5_K_M | 22.2 GB | 32 GB, 64 GB, 128 GB, 192 GB, 256 GB, 512 GBEstimate only; requires the complete GGUF quant group. | 1 file |
| KRAFTONKRAFTON/Raon-1-mini-preview-gguf ↗Observed 27 Oct 2026 | GGUF · Q6_K | 25.8 GB | 64 GB, 128 GB, 192 GB, 256 GB, 512 GBEstimate only; requires the complete GGUF quant group. | 1 file |
| KRAFTONKRAFTON/Raon-1-mini-preview-gguf ↗Observed 27 Oct 2026 | GGUF · Q8_0 | 30.7 GB | 64 GB, 128 GB, 192 GB, 256 GB, 512 GBEstimate only; requires the complete GGUF quant group. | 1 file |
CURATED TRACKING
Known issues
No curated issues verified for this snapshot. An empty list does not mean there are no bugs.
HARDWARE NOTES
We ran it on a Mac
We ran the Q4_K_M GGUF (19.69 GB, KRAFTON/Raon-1-mini-preview-gguf @2e6f93b) on an Apple M4 Max Mac, 128 GiB unified memory, macOS 27.0.1, Metal with krafton-ai/llama.cpp raon-1-mini-preview branch (fb89566); upstream llama.cpp support is pending on 26 Oct 2026.
Decode: 63.0 tok/s on a short prompt, 62.3 tok/s on a long prompt (medians of 3 runs). Prefill: 219 / 1,076 tok/s. Time to first token: 89 ms / 993 ms.
Peak memory: 20.0 GiB (llama-server footprint). Load time: 187.1 s.
llama-bench: pp512 1,239.4 tok/s, tg128 66.0 tok/s.
Thinking off, to match the protocol used for other models. The model card says Raon is designed for thinking mode, so these are no-think chat numbers.
One machine, one run set: temperature 0, 8K context, 1 warm-up plus 3 runs per prompt. Short prompt is 18 tokens; long prompt is 1,064 tokens (each model counts with its own tokenizer). Weights were loaded from an external USB drive with mmap off, so load time is disk-bound. Your speed will differ with hardware, context length and settings.
COMPARE
Compare with…
No comparison page is generated for this model yet. Browse all comparisons →
EMBED
Embed this badge
Works in GitHub READMEs and Hugging Face model cards. The badge updates when the site rebuilds.