# Can the Radeon RX 6700 XT run LFM3-VL-2B?

> Yes. The Radeon RX 6700 XT runs LFM3-VL-2B fully in VRAM at FP16 / BF16 (about 4.3 GB of 12 GB), at roughly 55 tokens/s.

Radeon RX 6700 XT: 12 GB (12 GB usable), 384 GB/s. Estimates at an 8K context.

## Every quantisation

| Quant | Quality | ≈ GB | Fit | Est. tok/s |
|---|---|---|---|---|
| Q2_K | Heavy loss | 1.56 | fits | 335 |
| Q3_K_M | Noticeable loss | 1.71 | fits | 225 |
| Q4_K_M | Small loss | 1.95 | fits | 182 |
| Q5_K_M | Near-lossless | 2.11 | fits | 155 |
| Q6_K | Near-lossless | 2.28 | fits | 134 |
| Q8_0 | Lossless | 2.69 | fits | 104 |
| F16 (pick) | Lossless | 4.27 | fits | 55 |

## Run it

From the Ollama library (ollama):

```bash
ollama run lfm3-vl:2b
```

F16 straight from Hugging Face (ollama):

```bash
ollama run hf.co/LiquidAI/LFM3-VL-2B-GGUF:F16
```

Docker with ROCm (AMD) (ollama):

```bash
docker run -d --name ollama --device /dev/kfd --device /dev/dri \
  -e HSA_OVERRIDE_GFX_VERSION=10.3.0 \
  -v ollama:/root/.ollama -p 11434:11434 ollama/ollama:rocm
docker exec -it ollama ollama run lfm3-vl:2b
```

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

```bash
llama-server -hf LiquidAI/LFM3-VL-2B-GGUF:F16
```

Docker with ROCm (AMD) (llama.cpp):

```bash
docker run -d --name llama --device /dev/kfd --device /dev/dri -p 8080:8080 \
  -e HSA_OVERRIDE_GFX_VERSION=10.3.0 \
  -v ~/.cache/llama.cpp:/root/.cache/llama.cpp \
  ghcr.io/ggml-org/llama.cpp:server-rocm \
  -hf LiquidAI/LFM3-VL-2B-GGUF:F16 --host 0.0.0.0 --port 8080 -ngl 99
```

Download with the lms CLI (lmstudio):

```bash
lms get LiquidAI/LFM3-VL-2B-GGUF@F16
```

Ask about an image (transformers):

```python
from transformers import pipeline

pipe = pipeline("image-text-to-text", model="LiquidAI/LFM3-VL-2B", 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/Qwen3.5-2B](https://selfhost.directory/models/qwen/qwen3.5-2b/on/rx-6700-xt): Q8_0 ≈2.84 GB, est. 87 tok/s
- [LiquidAI/d1-2B](https://selfhost.directory/models/liquidai/d1-2b/on/rx-6700-xt): Q8_0 ≈2.69 GB, est. 104 tok/s
- [Qwen/Qwen3-VL-2B-Instruct](https://selfhost.directory/models/qwen/qwen3-vl-2b-instruct/on/rx-6700-xt): Q8_0 ≈3.18 GB, est. 93 tok/s
- [openbmb/MiniCPM-V-4.7-1B](https://selfhost.directory/models/openbmb/minicpm-v-4.7-1b/on/rx-6700-xt): Q8_0 ≈1.94 GB, est. 128 tok/s
- [openbmb/MiniCPM-V-4.6](https://selfhost.directory/models/openbmb/minicpm-v-4.6/on/rx-6700-xt): Q8_0 ≈1.68 GB, est. 153 tok/s
- [HuggingFaceTB/SmolVLM2-2.2B-Instruct](https://selfhost.directory/models/huggingfacetb/smolvlm2-2.2b-instruct/on/rx-6700-xt): Q8_0 ≈2.58 GB, est. 88 tok/s

## Other GPUs that run it

- [GeForce RTX 4060](https://selfhost.directory/models/liquidai/lfm3-vl-2b/on/rtx-4060): F16, est. 39 tok/s
- [Apple M4 (16 GB)](https://selfhost.directory/models/liquidai/lfm3-vl-2b/on/apple-m4-16gb): F16, est. 17 tok/s
- [GeForce RTX 3060 12GB](https://selfhost.directory/models/liquidai/lfm3-vl-2b/on/rtx-3060): F16, est. 52 tok/s
- [Arc B580](https://selfhost.directory/models/liquidai/lfm3-vl-2b/on/arc-b580): F16, est. 65 tok/s
- [Tesla T4](https://selfhost.directory/models/liquidai/lfm3-vl-2b/on/tesla-t4): F16, est. 46 tok/s
- [Arc A770 16GB](https://selfhost.directory/models/liquidai/lfm3-vl-2b/on/arc-a770): F16, est. 80 tok/s
- [Radeon RX 7800 XT](https://selfhost.directory/models/liquidai/lfm3-vl-2b/on/rx-7800-xt): F16, est. 89 tok/s
- [Tesla P40](https://selfhost.directory/models/liquidai/lfm3-vl-2b/on/tesla-p40): F16, est. 50 tok/s

---
Model: https://selfhost.directory/models/liquidai/lfm3-vl-2b. GPU: https://selfhost.directory/models/gpus/rx-6700-xt. Source: https://selfhost.directory/models/liquidai/lfm3-vl-2b/on/rx-6700-xt