# Can the Tesla T4 run WeVisCut?

> Yes. The Tesla T4 runs WeVisCut fully in VRAM at Q8_0 (about 11 GB of 16 GB), at roughly 18 tokens/s.

Tesla T4: 16 GB (16 GB usable), 320 GB/s. Estimates at an 8K context.

## Every quantisation

| Quant | Quality | ≈ GB | Fit | Est. tok/s |
|---|---|---|---|---|
| Q2_K | Heavy loss | 4.53 | fits | 57 |
| Q3_K_M | Noticeable loss | 6.03 | fits | 38 |
| Q4_K_M | Small loss | 7 | fits | 31 |
| Q5_K_M | Near-lossless | 7.82 | fits | 26 |
| Q6_K | Near-lossless | 8.69 | fits | 23 |
| Q8_0 (pick) | Lossless | 10.81 | fits | 18 |
| F16 | Lossless | 19.02 | offload | 5.5 |

## Run it

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

```bash
llama-server -hf prithivMLmods/tencent-WeVisCut-GGUF:Q8_0
```

Serve an OpenAI-compatible API (vllm):

```bash
vllm serve tencent/WeVisCut
```

Docker Compose (NVIDIA) (vllm):

```yaml
services:
  vllm:
    image: vllm/vllm-openai:latest
    command: tencent/WeVisCut
    ports:
      - "8000:8000"
    ipc: host
    volumes:
      - ~/.cache/huggingface:/root/.cache/huggingface
    deploy:
      resources:
        reservations:
          devices:
            - driver: nvidia
              count: all
              capabilities: [gpu]
```

Serve an OpenAI-compatible API (sglang):

```bash
sglang serve tencent/WeVisCut --host 0.0.0.0 --port 30000
```

Docker Compose (NVIDIA) (sglang):

```yaml
services:
  sglang:
    image: lmsysorg/sglang:latest
    command: sglang serve tencent/WeVisCut --host 0.0.0.0 --port 30000
    ports:
      - "30000:30000"
    ipc: host
    volumes:
      - ~/.cache/huggingface:/root/.cache/huggingface
    deploy:
      resources:
        reservations:
          devices:
            - driver: nvidia
              count: all
              capabilities: [gpu]
```


## Omni models that fit comfortably

- [openbmb/MiniCPM-o-4_5](https://selfhost.directory/models/openbmb/minicpm-o-4_5/on/tesla-t4): Q8_0 ≈10.15 GB, est. 18 tok/s
- [google/gemma-4-E4B-it](https://selfhost.directory/models/google/gemma-4-e4b-it/on/tesla-t4): Q8_0 ≈9.17 GB, est. 21 tok/s
- [Qwen/Qwen2.5-Omni-7B](https://selfhost.directory/models/qwen/qwen2.5-omni-7b/on/tesla-t4): Q8_0 ≈8.86 GB, est. 15 tok/s
- [OpenMOSS-Team/MOSS-Music-8B-Instruct](https://selfhost.directory/models/openmoss-team/moss-music-8b-instruct/on/tesla-t4): Q8_0 ≈10.15 GB, est. 18 tok/s
- [google/gemma-4-12B-it](https://selfhost.directory/models/google/gemma-4-12b-it/on/tesla-t4): Q6_K ≈13.07 GB, est. 18 tok/s
- [nvidia/audio-flamingo-3-hf](https://selfhost.directory/models/nvidia/audio-flamingo-3-hf/on/tesla-t4): Q8_0 ≈10.05 GB, est. 20 tok/s

## Other GPUs that run it

- [GeForce RTX 4060](https://selfhost.directory/models/tencent/weviscut/on/rtx-4060): Q4_K_M, est. 26 tok/s
- [Apple M4 (16 GB)](https://selfhost.directory/models/tencent/weviscut/on/apple-m4-16gb): Q6_K, est. 8.5 tok/s
- [Radeon RX 6700 XT](https://selfhost.directory/models/tencent/weviscut/on/rx-6700-xt): Q8_0, est. 21 tok/s
- [GeForce RTX 3060 12GB](https://selfhost.directory/models/tencent/weviscut/on/rtx-3060): Q8_0, est. 20 tok/s
- [Arc B580](https://selfhost.directory/models/tencent/weviscut/on/arc-b580): Q8_0, est. 25 tok/s
- [RTX A4000](https://selfhost.directory/models/tencent/weviscut/on/rtx-a4000): Q8_0, est. 25 tok/s
- [Arc A770 16GB](https://selfhost.directory/models/tencent/weviscut/on/arc-a770): Q8_0, est. 31 tok/s
- [Radeon RX 7800 XT](https://selfhost.directory/models/tencent/weviscut/on/rx-7800-xt): Q8_0, est. 34 tok/s

---
Model: https://selfhost.directory/models/tencent/weviscut. GPU: https://selfhost.directory/models/gpus/tesla-t4. Source: https://selfhost.directory/models/tencent/weviscut/on/tesla-t4