selfhost.directory

The index of self-hostable projects — ranked, categorized and served with live updates.
Own your data and stop giving it to big tech.

KRAFTON logo

Can the Apple M4 (16 GB) run Raon-1-nano?

Raon-1-nano·3.5B·Apple M4 (16 GB)·16 GB unified memory

Saved in this browser, used on every model page.
Yes

Yes. The Apple M4 (16 GB) runs Raon-1-nano fully in unified memory at FP16 / BF16 (about 9.6 GB of 11 GB), at roughly 9.3 tokens/s.

best quant
FP16 / BF16
needs
9.6 GB
usable
11 GB
est. speed
~9.3 tok/s

Estimates for an 8K context; offload assumes 32 GB of system RAM.Try other contexts on the Raon-1-nano page →

Every quantisation of Raon-1-nano on the Apple M4 (16 GB)

QuantQualityMemoryFitSpeed
Q2_K Heavy loss4.6 GBfits~57 tok/s
Q3_K_M Noticeable loss5.2 GBfits~38 tok/s
Q4_K_M Small loss5.5 GBfits~31 tok/s
Q5_K_M Near-lossless6.0 GBfits~26 tok/s
Q6_K Near-lossless6.1 GBfits~23 tok/s
Q8_0 Lossless6.8 GBfits~18 tok/s
FP16 / BF16 pickLossless9.6 GBfits~9.3 tok/s

Speed is a bandwidth-based estimate for generating text (120 GB/s); prompt processing and long contexts are slower.

Measured speed on the Apple M4 (16 GB)

No community reports for this pair yet.

Report your own speed →

Run Raon-1-nano on the Apple M4 (16 GB)

Using FP16 / BF16, the best fit for Apple M4 (16 GB).

MLX · Apple Silicon with mlx-lmdocs
pip install mlx-lm
mlx_lm.generate --model KRAFTON/Raon-1-nano-mlx --prompt "Hello"

Use `mlx_lm.chat` with the same --model for an interactive chat.

MLX · OpenAI-compatible serverdocs
mlx_lm.server --model KRAFTON/Raon-1-nano-mlx --port 8080

API on http://localhost:8080/v1.

Ollama · F16 straight from Hugging Facedocs
ollama run hf.co/KRAFTON/Raon-1-nano-gguf:F16
llama.cpp · OpenAI-compatible server with a web UIdocs
llama-server -hf KRAFTON/Raon-1-nano-gguf:F16

brew install llama.cpp ships with Metal. Downloads the GGUF on first run.

LM Studio · Download with the lms CLIdocs
lms get KRAFTON/Raon-1-nano-gguf@F16

Then load it from the LM Studio app or with `lms load`.

Transformers · Chat with the pipeline APIdocs
from transformers import pipeline

pipe = pipeline("text-generation", model="KRAFTON/Raon-1-nano", dtype="auto", device_map="auto")
messages = [{"role": "user", "content": "Explain self-hosting in one paragraph."}]
out = pipe(messages, max_new_tokens=512)
print(out[0]["generated_text"][-1]["content"])

Chat models that fit the Apple M4 (16 GB) comfortably

All that fit →
ModelParamsBest quantMemorySpeedLicence
3.2BQ8_0≈4.7 GBFits in memory~19 tok/sConditions
4BQ8_0≈5.8 GBFits in memory~15 tok/sCommercial OK
3.1BQ8_0≈4.3 GBFits in memory~20 tok/sNon-commercial
3.8BQ8_0≈5.5 GBFits in memory~16 tok/sCommercial OK
ibm-granite logogranite-4.2-3bibm-granite
3.7BQ8_0≈4.9 GBFits in memory~17 tok/sCommercial OK
4BQ8_0≈5.8 GBFits in memory~15 tok/sCommercial OK

Other GPUs that run Raon-1-nano

Frequently asked questions

Can the Apple M4 (16 GB) run Raon-1-nano?
Yes. The Apple M4 (16 GB) runs Raon-1-nano fully in unified memory at FP16 / BF16 (about 9.6 GB of 11 GB), at roughly 9.3 tokens/s.
How fast does Raon-1-nano run on the Apple M4 (16 GB)?
An estimate from memory bandwidth gives about 9.3 tokens/s at FP16 / BF16.
Which quantisation of Raon-1-nano should I use on the Apple M4 (16 GB)?
FP16 / BF16: the largest one that fits. Expected quality: lossless compared with the full-precision weights.

Memory and speed are estimates from the model's size and the card's memory bandwidth; real numbers depend on the runtime, drivers and context length. Community figures are medians of user reports.