# Can the Apple M4 Max 40-core (128 GB) run Qwen3.8-Flash-Next?

> Yes, but only at Q3_K_M (about 87 GB of 96 GB), which costs noticeable quality, at roughly 10 tokens/s.

Apple M4 Max 40-core (128 GB): 128 GB (96 GB usable), 546 GB/s. Estimates at an 8K context.

## Every quantisation

| Quant | Quality | ≈ GB | Fit | Est. tok/s |
|---|---|---|---|---|
| Q2_K | Heavy loss | 59.12 | fits | 15 |
| Q3_K_M (pick) | Noticeable loss | 87.28 | fits | 10 |
| Q4_K_M | Small loss | 107.96 | no | — |
| Q5_K_M | Near-lossless | 126.44 | no | — |
| Q6_K | Near-lossless | 146.25 | no | — |
| Q8_0 | Lossless | 185.32 | no | — |
| F16 | Lossless | 347.46 | no | — |

## Run it

From the Ollama library (ollama):

```bash
ollama run qwen3.8-flash-next
```

UD-Q3_K_XL straight from Hugging Face (ollama):

```bash
ollama run hf.co/unsloth/Qwen3.8-Flash-Next-GGUF:UD-Q3_K_XL
```

OpenAI-compatible server with a web UI (llama.cpp):

```bash
llama-server -hf unsloth/Qwen3.8-Flash-Next-GGUF:UD-Q3_K_XL
```

Download with the lms CLI (lmstudio):

```bash
lms get unsloth/Qwen3.8-Flash-Next-GGUF@UD-Q3_K_XL
```

Ask about an image (transformers):

```python
from transformers import pipeline

pipe = pipeline("image-text-to-text", model="Qwen/Qwen3.8-Flash-Next", 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 comfortably

- [Qwen/Qwen3.5-122B-A10B](https://selfhost.directory/models/qwen/qwen3.5-122b-a10b/on/apple-m4-max-128gb-40c): Q5_K_M ≈90.74 GB, est. 42 tok/s
- [nvidia/NVIDIA-Nemotron-3.5-Super-VL-120B-A12B-BF16](https://selfhost.directory/models/nvidia/nvidia-nemotron-3.5-super-vl-120b-a12b-bf16/on/apple-m4-max-128gb-40c): Q5_K_M ≈88.31 GB, est. 35 tok/s
- [meta-llama/Llama-4-Scout-17B-16E-Instruct](https://selfhost.directory/models/meta-llama/llama-4-scout-17b-16e-instruct/on/apple-m4-max-128gb-40c): Q6_K ≈88.48 GB, est. 20 tok/s
- [zai-org/GLM-4.6V](https://selfhost.directory/models/zai-org/glm-4.6v/on/apple-m4-max-128gb-40c): Q5_K_M ≈80.9 GB, est. 30 tok/s
- [Qwen/Qwen2.5-VL-72B-Instruct](https://selfhost.directory/models/qwen/qwen2.5-vl-72b-instruct/on/apple-m4-max-128gb-40c): Q8_0 ≈79.27 GB, est. 3.9 tok/s
- [Qwen/Qwen3.6-35B-A3B](https://selfhost.directory/models/qwen/qwen3.6-35b-a3b/on/apple-m4-max-128gb-40c): Q8_0 ≈38.1 GB, est. 94 tok/s

## Other GPUs that run it

- [RTX PRO 6000 Blackwell](https://selfhost.directory/models/qwen/qwen3.8-flash-next/on/rtx-pro-6000-blackwell): Q3_K_M, est. 33 tok/s
- [Ryzen AI Max+ 395 (128 GB)](https://selfhost.directory/models/qwen/qwen3.8-flash-next/on/ryzen-ai-max-395-128gb): Q3_K_M, est. 4.7 tok/s
- [Apple M2 Ultra (192 GB)](https://selfhost.directory/models/qwen/qwen3.8-flash-next/on/apple-m2-ultra-192gb): Q5_K_M, est. 10 tok/s
- [Apple M3 Ultra (512 GB)](https://selfhost.directory/models/qwen/qwen3.8-flash-next/on/apple-m3-ultra-512gb): F16, est. 3.7 tok/s

---
Model: https://selfhost.directory/models/qwen/qwen3.8-flash-next. GPU: https://selfhost.directory/models/gpus/apple-m4-max-128gb-40c. Source: https://selfhost.directory/models/qwen/qwen3.8-flash-next/on/apple-m4-max-128gb-40c