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.

InternScience logo

Can the Apple M1 Max (64 GB) run MLExpert?

MLExpert·34.7B (3.5B active)·Apple M1 Max (64 GB)·64 GB unified memory

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

Yes. The Apple M1 Max (64 GB) runs MLExpert fully in unified memory at Q8_0 (about 37 GB of 48 GB), at roughly 60 tokens/s.

best quant
Q8_0
needs
37 GB
usable
48 GB
est. speed
~60 tok/s

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

Every quantisation of MLExpert on the Apple M1 Max (64 GB)

QuantQualityMemoryFitSpeed
Q2_K Heavy loss15 GBfits~194 tok/s
Q3_K_M Noticeable loss18 GBfits~130 tok/s
Q4_K_M Small loss23 GBfits~105 tok/s
Q5_K_M Near-lossless28 GBfits~90 tok/s
Q6_K Near-lossless31 GBfits~77 tok/s
Q8_0 pickLossless37 GBfits~60 tok/s
FP16 / BF16 Lossless69 GBtoo big—

Speed is a bandwidth-based estimate for generating text (400 GB/s); prompt processing and long contexts are slower.

Measured speed on the Apple M1 Max (64 GB)

No community reports for this pair yet.

Report your own speed →

Run MLExpert on the Apple M1 Max (64 GB)

Using Q8_0, the best fit for Apple M1 Max (64 GB).

Ollama · Q8_0 straight from Hugging Facedocs
ollama run hf.co/bartowski/InternScience_MLExpert-GGUF:Q8_0
llama.cpp · OpenAI-compatible server with a web UIdocs
llama-server -hf bartowski/InternScience_MLExpert-GGUF:Q8_0

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

LM Studio · Download with the lms CLIdocs
lms get bartowski/InternScience_MLExpert-GGUF@Q8_0

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="InternScience/MLExpert", 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 M1 Max (64 GB) comfortably

All that fit →
ModelParamsBest quantMemorySpeedLicence
36BA3BQ8_0≈38 GBFits in memory~69 tok/sCommercial OK
31.2BA4.6BQ8_0≈39 GBFits in memory~45 tok/sCommercial OK
32.8BQ8_0≈37 GBFits in memory~6.3 tok/sCommercial OK
RecursiveSI logoflintRecursiveSI
39.8BA1.3BQ8_0≈43 GBFits in memory~166 tok/sCommercial OK
20.9BA4.2BQ8_0≈13 GBFits in memory~49 tok/sCommercial OK
30.5BA3BQ8_0≈33 GBFits in memory~69 tok/sCommercial OK

Other GPUs that run MLExpert

Frequently asked questions

Can the Apple M1 Max (64 GB) run MLExpert?
Yes. The Apple M1 Max (64 GB) runs MLExpert fully in unified memory at Q8_0 (about 37 GB of 48 GB), at roughly 60 tokens/s.
How fast does MLExpert run on the Apple M1 Max (64 GB)?
An estimate from memory bandwidth gives about 60 tokens/s at Q8_0.
Which quantisation of MLExpert should I use on the Apple M1 Max (64 GB)?
Q8_0: 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.