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 M2 Ultra (192 GB) run Hy4-flash?

Hy4-flash·220B·Apple M2 Ultra (192 GB)·192 GB unified memory

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

Yes, but only at Q3_K_M (about 108 GB of 144 GB), which costs noticeable quality, at roughly 4.1 tokens/s.

best quant
Q3_K_M
needs
108 GB
usable
144 GB
est. speed
~4.1 tok/s

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

Every quantisation of Hy4-flash on the Apple M2 Ultra (192 GB)

QuantQualityMemoryFitSpeed
Q2_K Heavy loss74 GBfits~6.1 tok/s
Q3_K_M pickNoticeable loss108 GBfits~4.1 tok/s
Q4_K_M Small loss168 GBtoo big—
Q5_K_M Near-lossless156 GBtoo big—
Q6_K Near-lossless181 GBtoo big—
Q8_0 Lossless232 GBtoo big—
FP16 / BF16 Lossless434 GBtoo big—

Speed is a bandwidth-based estimate for generating text (800 GB/s); prompt processing and long contexts are slower.

Measured speed on the Apple M2 Ultra (192 GB)

No community reports for this pair yet.

Report your own speed →

Run Hy4-flash on the Apple M2 Ultra (192 GB)

Transformers · Ask about an imagedocs
from transformers import pipeline

pipe = pipeline("image-text-to-text", model="tencent/Hy4-flash", 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 M2 Ultra (192 GB) comfortably

All that fit →
ModelParamsBest quantMemorySpeedLicence
180BA61.6BQ5_K_M≈126 GBFits in memory~10 tok/sConditions
220BQ8_0≈3.5 GBFits in memory~1.9 tok/sConditions
125BA10BQ8_0≈128 GBFits in memory~41 tok/sCommercial OK
stepfun-ai logoStep-3.7-Flashstepfun-ai
201BQ4_K_M≈131 GBFits in memory~3.6 tok/sCommercial OK
124BA12BQ8_0≈130 GBFits in memory~35 tok/sCommercial OK
109BA18BQ8_0≈114 GBFits in memory~23 tok/sConditions

Other GPUs that run Hy4-flash

Frequently asked questions

Can the Apple M2 Ultra (192 GB) run Hy4-flash?
Yes, but only at Q3_K_M (about 108 GB of 144 GB), which costs noticeable quality, at roughly 4.1 tokens/s.
How fast does Hy4-flash run on the Apple M2 Ultra (192 GB)?
An estimate from memory bandwidth gives about 4.1 tokens/s at Q3_K_M.
Which quantisation of Hy4-flash should I use on the Apple M2 Ultra (192 GB)?
Q3_K_M: the largest one that fits. Expected quality: noticeable loss 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.