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.

tencent logo

Can the Apple M3 Ultra (512 GB) run Hy4-flash-base?

Hy4-flash-base·219B·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 Hy4-flash-base fully in unified memory at Q8_0 (about 231 GB of 384 GB), at roughly 1.9 tokens/s.

best quant
Q8_0
needs
231 GB
usable
384 GB
est. speed
~1.9 tok/s

Estimates for an 8K context; offload assumes 32 GB of system RAM.Try other contexts on the Hy4-flash-base page →

Every quantisation of Hy4-flash-base on the Apple M3 Ultra (512 GB)

QuantQualityMemoryFitSpeed
Q2_K Heavy loss74 GBfits~6.2 tok/s
Q3_K_M Noticeable loss108 GBfits~4.2 tok/s
Q4_K_M Small loss133 GBfits~3.4 tok/s
Q5_K_M Near-lossless156 GBfits~2.9 tok/s
Q6_K Near-lossless180 GBfits~2.5 tok/s
Q8_0 pickLossless231 GBfits~1.9 tok/s
FP16 / BF16 Lossless432 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 Hy4-flash-base on the Apple M3 Ultra (512 GB)

Transformers · Ask about an imagedocs
from transformers import pipeline

pipe = pipeline("image-text-to-text", model="tencent/Hy4-flash-base", 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
180BA61.6BQ8_0≈185 GBFits in memory~6.9 tok/sConditions
220BQ8_0≈3.5 GBFits in memory~1.9 tok/sConditions
321BA18.3BQ8_0≈356 GBFits in memory~23 tok/sCommercial OK
236BA22BQ8_0≈246 GBFits in memory~19 tok/sCommercial OK
thinkingmachines logoInkling-Smallthinkingmachines
266BA1.7BQ8_0≈276 GBFits in memory~247 tok/sCommercial OK
125BA10BQ8_0≈128 GBFits in memory~42 tok/sCommercial OK

Other GPUs that run Hy4-flash-base

Frequently asked questions

Can the Apple M3 Ultra (512 GB) run Hy4-flash-base?
Yes. The Apple M3 Ultra (512 GB) runs Hy4-flash-base fully in unified memory at Q8_0 (about 231 GB of 384 GB), at roughly 1.9 tokens/s.
How fast does Hy4-flash-base run on the Apple M3 Ultra (512 GB)?
An estimate from memory bandwidth gives about 1.9 tokens/s at Q8_0.
Which quantisation of Hy4-flash-base 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.