# Can the Arc A770 16GB run plamo-3-8b-fin-instruct?

> Yes. The Arc A770 16GB runs plamo-3-8b-fin-instruct fully in VRAM at Q8_0 (about 9.9 GB of 16 GB), at roughly 34 tokens/s.

Arc A770 16GB: 16 GB (16 GB usable), 560 GB/s. Estimates at an 8K context.

## Every quantisation

| Quant | Quality | ≈ GB | Fit | Est. tok/s |
|---|---|---|---|---|
| Q2_K | Heavy loss | 4.37 | fits | 110 |
| Q3_K_M | Noticeable loss | 5.21 | fits | 74 |
| Q4_K_M | Small loss | 6.17 | fits | 60 |
| Q5_K_M | Near-lossless | 6.99 | fits | 51 |
| Q6_K | Near-lossless | 7.85 | fits | 44 |
| Q8_0 (pick) | Lossless | 9.87 | fits | 34 |
| F16 | Lossless | 17.7 | offload | 10 |

## Run it

Q8_0 straight from Hugging Face (ollama):

```bash
ollama run hf.co/mradermacher/plamo-3-8b-fin-instruct-GGUF:Q8_0
```

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

```bash
llama-server -hf mradermacher/plamo-3-8b-fin-instruct-GGUF:Q8_0
```

Docker with Vulkan (llama.cpp):

```bash
docker run -d --name llama --device /dev/dri -p 8080:8080 \
  -v ~/.cache/llama.cpp:/root/.cache/llama.cpp \
  ghcr.io/ggml-org/llama.cpp:server-vulkan \
  -hf mradermacher/plamo-3-8b-fin-instruct-GGUF:Q8_0 --host 0.0.0.0 --port 8080 -ngl 99
```

Download with the lms CLI (lmstudio):

```bash
lms get mradermacher/plamo-3-8b-fin-instruct-GGUF@Q8_0
```

Chat with the pipeline API (transformers):

```python
from transformers import pipeline

pipe = pipeline("text-generation", model="pfnet/plamo-3-8b-fin-instruct", dtype="auto", device_map="auto")
messages = [{"role": "user", "content": "Explain self-hosting in one paragraph."}]
out = pipe(messages, max_new_tokens=512)
print(out[0]["generated_text"][-1]["content"])
```


## Chat models that fit comfortably

- [Qwen/Qwen3-8B](https://selfhost.directory/models/qwen/qwen3-8b/on/arc-a770): Q8_0 ≈10.15 GB, est. 35 tok/s
- [meta-llama/Llama-3.1-8B-Instruct](https://selfhost.directory/models/meta-llama/llama-3.1-8b-instruct/on/arc-a770): Q8_0 ≈9.84 GB, est. 36 tok/s
- [meta-llama/Meta-Llama-3-8B-Instruct](https://selfhost.directory/models/meta-llama/meta-llama-3-8b-instruct/on/arc-a770): Q8_0 ≈9.85 GB, est. 36 tok/s
- [Qwen/Qwen2.5-7B-Instruct](https://selfhost.directory/models/qwen/qwen2.5-7b-instruct/on/arc-a770): Q8_0 ≈8.86 GB, est. 38 tok/s
- [xtools-at/DeepSeek-V4-Pro-Qwen3.5-9B-UNCENSORED-heretic](https://selfhost.directory/models/xtools-at/deepseek-v4-pro-qwen3.5-9b-uncensored-heretic/on/arc-a770): Q8_0 ≈11.07 GB, est. 31 tok/s
- [google/gemma-2-9b-it](https://selfhost.directory/models/google/gemma-2-9b-it/on/arc-a770): Q8_0 ≈12.74 GB, est. 31 tok/s

## Other GPUs that run it

- [GeForce RTX 4060](https://selfhost.directory/models/pfnet/plamo-3-8b-fin-instruct/on/rtx-4060): Q5_K_M, est. 25 tok/s
- [Apple M4 (16 GB)](https://selfhost.directory/models/pfnet/plamo-3-8b-fin-instruct/on/apple-m4-16gb): Q8_0, est. 7.3 tok/s
- [Radeon RX 6700 XT](https://selfhost.directory/models/pfnet/plamo-3-8b-fin-instruct/on/rx-6700-xt): Q8_0, est. 23 tok/s
- [GeForce RTX 3060 12GB](https://selfhost.directory/models/pfnet/plamo-3-8b-fin-instruct/on/rtx-3060): Q8_0, est. 22 tok/s
- [Arc B580](https://selfhost.directory/models/pfnet/plamo-3-8b-fin-instruct/on/arc-b580): Q8_0, est. 28 tok/s
- [Tesla T4](https://selfhost.directory/models/pfnet/plamo-3-8b-fin-instruct/on/tesla-t4): Q8_0, est. 19 tok/s
- [Radeon RX 7800 XT](https://selfhost.directory/models/pfnet/plamo-3-8b-fin-instruct/on/rx-7800-xt): Q8_0, est. 38 tok/s
- [Tesla P40](https://selfhost.directory/models/pfnet/plamo-3-8b-fin-instruct/on/tesla-p40): F16, est. 11 tok/s

---
Model: https://selfhost.directory/models/pfnet/plamo-3-8b-fin-instruct. GPU: https://selfhost.directory/models/gpus/arc-a770. Source: https://selfhost.directory/models/pfnet/plamo-3-8b-fin-instruct/on/arc-a770