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.

nex-agi logo

Can the Apple M3 Ultra (512 GB) run Nex-N2.6?

Nex-N2.6·321B (18.3B active)·Apple M3 Ultra (512 GB)·512 GB unified memory

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

Yes. The Apple M3 Ultra (512 GB) runs Nex-N2.6 fully in unified memory at Q8_0 (about 357 GB of 384 GB), at roughly 23 tokens/s.

best quant
Q8_0
needs
357 GB
usable
384 GB
est. speed
~23 tok/s

Estimates for an 8K context; offload assumes 32 GB of system RAM.Try other contexts on the Nex-N2.6 page →

Every quantisation of Nex-N2.6 on the Apple M3 Ultra (512 GB)

QuantQualityMemoryFitSpeed
Q2_K Heavy loss146 GBfits~75 tok/s
Q3_K_M Noticeable loss175 GBfits~50 tok/s
Q4_K_M Small loss219 GBfits~41 tok/s
Q5_K_M Near-lossless261 GBfits~35 tok/s
Q6_K Near-lossless298 GBfits~30 tok/s
Q8_0 pickLossless357 GBfits~23 tok/s
FP16 / BF16 Lossless651 GBtoo big—

Speed is a bandwidth-based estimate for generating text (819 GB/s); prompt processing and long contexts are slower.

Measured speed on the Apple M3 Ultra (512 GB)

No community reports for this pair yet.

Report your own speed →

Run Nex-N2.6 on the Apple M3 Ultra (512 GB)

Using Q8_0, the best fit for Apple M3 Ultra (512 GB).

Ollama · Q8_0 straight from Hugging Facedocs
ollama run hf.co/bartowski/Nex-N2.6-GGUF:Q8_0
llama.cpp · OpenAI-compatible server with a web UIdocs
llama-server -hf bartowski/Nex-N2.6-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/Nex-N2.6-GGUF@Q8_0

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

Transformers · Ask about an imagedocs
from transformers import pipeline

pipe = pipeline("image-text-to-text", model="nex-agi/Nex-N2.6", dtype="auto", device_map="auto")
messages = [
    {
        "role": "user",
        "content": [
            {"type": "image", "url": "https://huggingface.co/datasets/huggingface/documentation-images/resolve/main/bee.jpg"},
            {"type": "text", "text": "Describe this image."},
        ],
    }
]
out = pipe(text=messages, max_new_tokens=256)
print(out[0]["generated_text"][-1]["content"])

Vision models that fit the Apple M3 Ultra (512 GB) comfortably

All that fit →
ModelParamsBest quantMemorySpeedLicence
321BA18.3BQ8_0≈356 GBFits in memory~23 tok/sCommercial OK
180BA61.6BQ8_0≈185 GBFits in memory~6.9 tok/sConditions
MiniMaxAI logoMiniMax-M3MiniMaxAI
427BQ6_K≈345 GBFits in memory~1.3 tok/sConditions
220BQ8_0≈3.5 GBFits in memory~1.9 tok/sConditions
thinkingmachines logoInkling-2-Smallthinkingmachines
383BQ6_K≈311 GBFits in memory~1.4 tok/sCommercial OK
thinkingmachines logoInkling-Smallthinkingmachines
266BA1.7BQ8_0≈276 GBFits in memory~247 tok/sCommercial OK

Frequently asked questions

Can the Apple M3 Ultra (512 GB) run Nex-N2.6?
Yes. The Apple M3 Ultra (512 GB) runs Nex-N2.6 fully in unified memory at Q8_0 (about 357 GB of 384 GB), at roughly 23 tokens/s.
How fast does Nex-N2.6 run on the Apple M3 Ultra (512 GB)?
An estimate from memory bandwidth gives about 23 tokens/s at Q8_0.
Which quantisation of Nex-N2.6 should I use on the Apple M3 Ultra (512 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.