Can the Apple M3 Ultra (512 GB) run Agnes-2.5-Pro-Theta?
Agnes-2.5-Pro-Theta·188B·Apple M3 Ultra (512 GB)·512 GB unified memory
Yes. The Apple M3 Ultra (512 GB) runs Agnes-2.5-Pro-Theta fully in unified memory at Q8_0 (about 201 GB of 384 GB), at roughly 2.3 tokens/s.
- best quant
- Q8_0
- needs
- 201 GB
- usable
- 384 GB
- est. speed
- ~2.3 tok/s
Estimates for an 8K context; offload assumes 32 GB of system RAM.Try other contexts on the Agnes-2.5-Pro-Theta page →
Every quantisation of Agnes-2.5-Pro-Theta on the Apple M3 Ultra (512 GB)
| Quant | Memory | Fit | Speed |
|---|---|---|---|
| Q2_K | 66 GB | fits | ~7.3 tok/s |
| Q3_K_M | 95 GB | fits | ~4.9 tok/s |
| Q4_K_M | 117 GB | fits | ~4 tok/s |
| Q5_K_M | 136 GB | fits | ~3.4 tok/s |
| Q6_K | 157 GB | fits | ~2.9 tok/s |
| Q8_0 pick | 201 GB | fits | ~2.3 tok/s |
| FP16 / BF16 | 373 GB | tight | ~1.2 tok/s |
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 Agnes-2.5-Pro-Theta on the Apple M3 Ultra (512 GB)
from transformers import pipeline
pipe = pipeline("image-text-to-text", model="Agnes-AI/Agnes-2.5-Pro-Theta", 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 →| Model | Params | Best quant | Memory | Speed |
|---|---|---|---|---|
| 180BA61.6B | Q8_0 | ≈185 GBFits in memory | ~6.9 tok/s | |
| 220B | Q8_0 | ≈3.5 GBFits in memory | ~1.9 tok/s | |
| 321BA18.3B | Q8_0 | ≈356 GBFits in memory | ~23 tok/s | |
| 125BA10B | Q8_0 | ≈128 GBFits in memory | ~42 tok/s | |
| 236BA22B | Q8_0 | ≈246 GBFits in memory | ~19 tok/s | |
| 266BA1.7B | Q8_0 | ≈276 GBFits in memory | ~247 tok/s |
Other GPUs that run Agnes-2.5-Pro-Theta
Frequently asked questions
- Can the Apple M3 Ultra (512 GB) run Agnes-2.5-Pro-Theta?
- Yes. The Apple M3 Ultra (512 GB) runs Agnes-2.5-Pro-Theta fully in unified memory at Q8_0 (about 201 GB of 384 GB), at roughly 2.3 tokens/s.
- How fast does Agnes-2.5-Pro-Theta run on the Apple M3 Ultra (512 GB)?
- An estimate from memory bandwidth gives about 2.3 tokens/s at Q8_0.
- Which quantisation of Agnes-2.5-Pro-Theta 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.