Multimodal Models
111 profiles in this group. Use this hub page to compare practical VRAM floor, expected throughput, and best local-vs-cloud path.
| Model | Data | VRAM min | VRAM optimal | Best local GPU | Cloud fallback | Detail |
|---|---|---|---|---|---|---|
| Llama 4 128X17B FP16 | Estimated | 430GB | 442GB | Cloud-first (no practical single-GPU local) | H100/H200 class | Open |
| Llama 4 128X17B Q4 | Estimated | 418GB | 428GB | Cloud-first (no practical single-GPU local) | H100/H200 class | Open |
| Llama 4 128X17B Q5 | Estimated | 420GB | 430GB | Cloud-first (no practical single-GPU local) | H100/H200 class | Open |
| Llama 4 128X17B Q8 | Estimated | 424GB | 434GB | Cloud-first (no practical single-GPU local) | H100/H200 class | Open |
| Qwen3.5 397B-A17B CLOUD | Estimated | 215GB | 223GB | Cloud-first (no practical single-GPU local) | H100/H200 class | Open |
| Llama 4 16X17B FP16 | Estimated | 225GB | 237GB | Cloud-first (no practical single-GPU local) | H100/H200 class | Open |
| Llama 4 16X17B Q4 | Estimated | 213GB | 223GB | Cloud-first (no practical single-GPU local) | H100/H200 class | Open |
| Llama 4 16X17B Q5 | Estimated | 215GB | 225GB | Cloud-first (no practical single-GPU local) | H100/H200 class | Open |
| Llama 4 16X17B Q8 | Estimated | 219GB | 229GB | Cloud-first (no practical single-GPU local) | H100/H200 class | Open |
| Qwen3 VL 235B CLOUD | Estimated | 140GB | 148GB | Cloud-first (no practical single-GPU local) | H100/H200 class | Open |
| Qwen3 VL 235B FP16 | Estimated | 150GB | 162GB | Cloud-first (no practical single-GPU local) | H100/H200 class | Open |
| Qwen3 VL 235B Q4 | Estimated | 138GB | 148GB | Cloud-first (no practical single-GPU local) | H100/H200 class | Open |
| Qwen3 VL 235B Q5 | Estimated | 140GB | 150GB | Cloud-first (no practical single-GPU local) | H100/H200 class | Open |
| Qwen3 VL 235B Q8 | Estimated | 144GB | 154GB | Cloud-first (no practical single-GPU local) | H100/H200 class | Open |
| Llama 3.2 Vision 90B FP16 | Estimated | 80GB | 92GB | Cloud-first (no practical single-GPU local) | H100/H200 class | Open |
| Llama 3.2 Vision 90B Q4 | Estimated | 68GB | 78GB | Dual RTX 4090 (model parallel) | A100 80GB | Open |
| Llama 3.2 Vision 90B Q5 | Estimated | 70GB | 80GB | Dual RTX 4090 (model parallel) | A100 80GB | Open |
| Llama 3.2 Vision 90B Q8 | Estimated | 74GB | 84GB | Cloud-first (no practical single-GPU local) | H100/H200 class | Open |
| Qwen2.5 VL 72B FP16 | Estimated | 50GB | 62GB | Dual RTX 4090 (model parallel) | A100 80GB | Open |
| Qwen2.5 VL 72B Q4 | Estimated | 38GB | 48GB | RTX 6000 Ada 48GB | A100 80GB | Open |
| Qwen2.5 VL 72B Q5 | Estimated | 40GB | 50GB | Dual RTX 4090 (model parallel) | A100 80GB | Open |
| Qwen2.5 VL 72B Q8 | Estimated | 44GB | 54GB | Dual RTX 4090 (model parallel) | A100 80GB | Open |
| LLaVA 34B FP16 | Estimated | 30GB | 42GB | RTX 6000 Ada 48GB | A100 80GB | Open |
| LLaVA 34B Q4 | Estimated | 18GB | 28GB | RTX 6000 Ada 48GB | A100 80GB | Open |
| LLaVA 34B Q5 | Estimated | 20GB | 30GB | RTX 6000 Ada 48GB | A100 80GB | Open |
| LLaVA 34B Q8 | Estimated | 24GB | 34GB | RTX 6000 Ada 48GB | A100 80GB | Open |
| Qwen2.5 VL 32B FP16 | Estimated | 30GB | 42GB | RTX 6000 Ada 48GB | A100 80GB | Open |
| Qwen2.5 VL 32B Q4 | Estimated | 18GB | 28GB | RTX 6000 Ada 48GB | A100 80GB | Open |
| Qwen2.5 VL 32B Q5 | Estimated | 20GB | 30GB | RTX 6000 Ada 48GB | A100 80GB | Open |
| Qwen2.5 VL 32B Q8 | Estimated | 24GB | 34GB | RTX 6000 Ada 48GB | A100 80GB | Open |
| Qwen3 VL 32B CLOUD | Estimated | 20GB | 28GB | RTX 6000 Ada 48GB | A100 80GB | Open |
| Qwen3 VL 32B FP16 | Estimated | 30GB | 42GB | RTX 6000 Ada 48GB | A100 80GB | Open |
| Qwen3 VL 32B Q4 | Estimated | 18GB | 28GB | RTX 6000 Ada 48GB | A100 80GB | Open |
| Qwen3 VL 32B Q5 | Estimated | 20GB | 30GB | RTX 6000 Ada 48GB | A100 80GB | Open |
| Qwen3 VL 32B Q8 | Estimated | 24GB | 34GB | RTX 6000 Ada 48GB | A100 80GB | Open |
| Qwen3 VL 30B CLOUD | Estimated | 20GB | 28GB | RTX 6000 Ada 48GB | A100 80GB | Open |
| Qwen3 VL 30B FP16 | Estimated | 30GB | 42GB | RTX 6000 Ada 48GB | A100 80GB | Open |
| Qwen3 VL 30B Q4 | Estimated | 18GB | 28GB | RTX 6000 Ada 48GB | A100 80GB | Open |
| Qwen3 VL 30B Q5 | Estimated | 20GB | 30GB | RTX 6000 Ada 48GB | A100 80GB | Open |
| Qwen3 VL 30B Q8 | Estimated | 24GB | 34GB | RTX 6000 Ada 48GB | A100 80GB | Open |
| Gemma 3 27B FP16 | Estimated | 30GB | 42GB | RTX 6000 Ada 48GB | A100 80GB | Open |
| Gemma 3 27B Q4 | Estimated | 18GB | 28GB | RTX 6000 Ada 48GB | A100 80GB | Open |
| Gemma 3 27B Q5 | Estimated | 20GB | 30GB | RTX 6000 Ada 48GB | A100 80GB | Open |
| Gemma 3 27B Q8 | Estimated | 24GB | 34GB | RTX 6000 Ada 48GB | A100 80GB | Open |
| LLaVA 13B FP16 | Estimated | 22GB | 34GB | RTX 6000 Ada 48GB | A100 80GB | Open |
| LLaVA 13B Q4 | Estimated | 10GB | 20GB | RTX 3090 24GB | A6000 48GB | Open |
| LLaVA 13B Q5 | Estimated | 12GB | 22GB | RTX 3090 24GB | A6000 48GB | Open |
| LLaVA 13B Q8 | Estimated | 16GB | 26GB | RTX 6000 Ada 48GB | A100 80GB | Open |
| Gemma 3 12B FP16 | Estimated | 22GB | 34GB | RTX 6000 Ada 48GB | A100 80GB | Open |
| Gemma 3 12B Q4 | Estimated | 10GB | 20GB | RTX 3090 24GB | A6000 48GB | Open |
| Gemma 3 12B Q5 | Estimated | 12GB | 22GB | RTX 3090 24GB | A6000 48GB | Open |
| Gemma 3 12B Q8 | Estimated | 16GB | 26GB | RTX 6000 Ada 48GB | A100 80GB | Open |
| Llama 3.2 Vision 11B FP16 | Estimated | 22GB | 34GB | RTX 6000 Ada 48GB | A100 80GB | Open |
| Llama 3.2 Vision 11B Q4 | Estimated | 12GB | 14GB | RTX 3090 24GB | A6000 48GB | Open |
| Llama 3.2 Vision 11B Q5 | Estimated | 14GB | 16GB | RTX 3090 24GB | A6000 48GB | Open |
| Llama 3.2 Vision 11B Q8 | Estimated | 16GB | 26GB | RTX 6000 Ada 48GB | A100 80GB | Open |
| LLaVA Llama3 8B FP16 | Estimated | 18GB | 30GB | RTX 6000 Ada 48GB | A100 80GB | Open |
| LLaVA Llama3 8B Q4 | Estimated | 6GB | 16GB | RTX 3090 24GB | A6000 48GB | Open |
| LLaVA Llama3 8B Q5 | Estimated | 8GB | 18GB | RTX 3090 24GB | A6000 48GB | Open |
| LLaVA Llama3 8B Q8 | Estimated | 12GB | 22GB | RTX 3090 24GB | A6000 48GB | Open |
| MiniCPM-V 8B FP16 | Estimated | 18GB | 30GB | RTX 6000 Ada 48GB | A100 80GB | Open |
| MiniCPM-V 8B Q4 | Estimated | 6GB | 16GB | RTX 3090 24GB | A6000 48GB | Open |
| MiniCPM-V 8B Q5 | Estimated | 8GB | 18GB | RTX 3090 24GB | A6000 48GB | Open |
| MiniCPM-V 8B Q8 | Estimated | 12GB | 22GB | RTX 3090 24GB | A6000 48GB | Open |
| Qwen3 VL 8B CLOUD | Estimated | 8GB | 16GB | RTX 3090 24GB | A6000 48GB | Open |
| Qwen3 VL 8B FP16 | Estimated | 18GB | 30GB | RTX 6000 Ada 48GB | A100 80GB | Open |
| Qwen3 VL 8B Q4 | Estimated | 6GB | 16GB | RTX 3090 24GB | A6000 48GB | Open |
| Qwen3 VL 8B Q5 | Estimated | 8GB | 18GB | RTX 3090 24GB | A6000 48GB | Open |
| Qwen3 VL 8B Q8 | Estimated | 12GB | 22GB | RTX 3090 24GB | A6000 48GB | Open |
| LLaVA 7B FP16 | Estimated | 18GB | 30GB | RTX 6000 Ada 48GB | A100 80GB | Open |
| LLaVA 7B Q4 | Estimated | 8GB | 10GB | RTX 3090 24GB | A6000 48GB | Open |
| LLaVA 7B Q5 | Estimated | 10GB | 12GB | RTX 3090 24GB | A6000 48GB | Open |
| LLaVA 7B Q8 | Estimated | 12GB | 22GB | RTX 3090 24GB | A6000 48GB | Open |
| Qwen2.5 VL 7B FP16 | Estimated | 18GB | 30GB | RTX 6000 Ada 48GB | A100 80GB | Open |
| Qwen2.5 VL 7B Q4 | Estimated | 6GB | 16GB | RTX 3090 24GB | A6000 48GB | Open |
| Qwen2.5 VL 7B Q5 | Estimated | 8GB | 18GB | RTX 3090 24GB | A6000 48GB | Open |
| Qwen2.5 VL 7B Q8 | Estimated | 12GB | 22GB | RTX 3090 24GB | A6000 48GB | Open |
| Gemma 3 4B FP16 | Estimated | 16GB | 28GB | RTX 6000 Ada 48GB | A100 80GB | Open |
| Gemma 3 4B Q4 | Estimated | 4GB | 14GB | RTX 3090 24GB | A6000 48GB | Open |
| Gemma 3 4B Q5 | Estimated | 6GB | 16GB | RTX 3090 24GB | A6000 48GB | Open |
| Gemma 3 4B Q8 | Estimated | 10GB | 20GB | RTX 3090 24GB | A6000 48GB | Open |
| Gemma 3n E4B FP16 | Estimated | 16GB | 28GB | RTX 6000 Ada 48GB | A100 80GB | Open |
| Gemma 3n E4B Q4 | Estimated | 4GB | 14GB | RTX 3090 24GB | A6000 48GB | Open |
| Gemma 3n E4B Q5 | Estimated | 6GB | 16GB | RTX 3090 24GB | A6000 48GB | Open |
| Gemma 3n E4B Q8 | Estimated | 10GB | 20GB | RTX 3090 24GB | A6000 48GB | Open |
| Qwen3 VL 4B CLOUD | Estimated | 6GB | 14GB | RTX 3090 24GB | A6000 48GB | Open |
| Qwen3 VL 4B FP16 | Estimated | 16GB | 28GB | RTX 6000 Ada 48GB | A100 80GB | Open |
| Qwen3 VL 4B Q4 | Estimated | 4GB | 14GB | RTX 3090 24GB | A6000 48GB | Open |
| Qwen3 VL 4B Q5 | Estimated | 6GB | 16GB | RTX 3090 24GB | A6000 48GB | Open |
| Qwen3 VL 4B Q8 | Estimated | 10GB | 20GB | RTX 3090 24GB | A6000 48GB | Open |
| Qwen2.5 VL 3B FP16 | Estimated | 16GB | 28GB | RTX 6000 Ada 48GB | A100 80GB | Open |
| Qwen2.5 VL 3B Q4 | Estimated | 4GB | 14GB | RTX 3090 24GB | A6000 48GB | Open |
| Qwen2.5 VL 3B Q5 | Estimated | 6GB | 16GB | RTX 3090 24GB | A6000 48GB | Open |
| Qwen2.5 VL 3B Q8 | Estimated | 10GB | 20GB | RTX 3090 24GB | A6000 48GB | Open |
| Gemma 3n E2B FP16 | Estimated | 14GB | 26GB | RTX 6000 Ada 48GB | A100 80GB | Open |
| Gemma 3n E2B Q4 | Estimated | 2GB | 12GB | RTX 3090 24GB | A6000 48GB | Open |
| Gemma 3n E2B Q5 | Estimated | 4GB | 14GB | RTX 3090 24GB | A6000 48GB | Open |
| Gemma 3n E2B Q8 | Estimated | 8GB | 18GB | RTX 3090 24GB | A6000 48GB | Open |
| Qwen3 VL 2B CLOUD | Estimated | 4GB | 12GB | RTX 3090 24GB | A6000 48GB | Open |
| Qwen3 VL 2B FP16 | Estimated | 14GB | 26GB | RTX 6000 Ada 48GB | A100 80GB | Open |
| Qwen3 VL 2B Q4 | Estimated | 2GB | 12GB | RTX 3090 24GB | A6000 48GB | Open |
| Qwen3 VL 2B Q5 | Estimated | 4GB | 14GB | RTX 3090 24GB | A6000 48GB | Open |
| Qwen3 VL 2B Q8 | Estimated | 8GB | 18GB | RTX 3090 24GB | A6000 48GB | Open |
| Gemma 3 1B FP16 | Estimated | 14GB | 26GB | RTX 6000 Ada 48GB | A100 80GB | Open |
| Gemma 3 1B Q4 | Estimated | 2GB | 12GB | RTX 3090 24GB | A6000 48GB | Open |
| Gemma 3 1B Q5 | Estimated | 4GB | 14GB | RTX 3090 24GB | A6000 48GB | Open |
| Gemma 3 1B Q8 | Estimated | 8GB | 18GB | RTX 3090 24GB | A6000 48GB | Open |
| Gemma 3 270M FP16 | Estimated | 12GB | 24GB | RTX 3090 24GB | A6000 48GB | Open |
| Gemma 3 270M Q4 | Estimated | 2GB | 10GB | RTX 3090 24GB | A6000 48GB | Open |
| Gemma 3 270M Q5 | Estimated | 2GB | 12GB | RTX 3090 24GB | A6000 48GB | Open |
| Gemma 3 270M Q8 | Estimated | 6GB | 16GB | RTX 3090 24GB | A6000 48GB | Open |
We may earn a commission if you click links on this page.