Reasoning Models
102 profiles in this group. Use this hub page to compare practical VRAM floor, expected throughput, and best local-vs-cloud path.
RAM planning is included because Ollama may spill layers to CPU when VRAM is tight, especially with long context.
| Model | Data | VRAM min | VRAM optimal | System RAM | RTX 3090 fit | Best local GPU | Cloud fallback | Detail |
|---|---|---|---|---|---|---|---|---|
| DeepSeek-R1 671B FP16 | Estimated | 430GB | 442GB | 128GB+ | Cloud / larger GPU | Cloud-first (no practical single-GPU local) | H100/H200 class | Open |
| DeepSeek-R1 671B Q4 | Estimated | 418GB | 428GB | 128GB+ | Cloud / larger GPU | Cloud-first (no practical single-GPU local) | H100/H200 class | Open |
| DeepSeek-R1 671B Q5 | Estimated | 420GB | 430GB | 128GB+ | Cloud / larger GPU | Cloud-first (no practical single-GPU local) | H100/H200 class | Open |
| DeepSeek-R1 671B Q8 | Estimated | 424GB | 434GB | 128GB+ | Cloud / larger GPU | Cloud-first (no practical single-GPU local) | H100/H200 class | Open |
| DeepSeek-V3 671B FP16 | Estimated | 430GB | 442GB | 128GB+ | Cloud / larger GPU | Cloud-first (no practical single-GPU local) | H100/H200 class | Open |
| DeepSeek-V3 671B Q4 | Estimated | 418GB | 428GB | 128GB+ | Cloud / larger GPU | Cloud-first (no practical single-GPU local) | H100/H200 class | Open |
| DeepSeek-V3 671B Q5 | Estimated | 420GB | 430GB | 128GB+ | Cloud / larger GPU | Cloud-first (no practical single-GPU local) | H100/H200 class | Open |
| DeepSeek-V3 671B Q8 | Estimated | 424GB | 434GB | 128GB+ | Cloud / larger GPU | Cloud-first (no practical single-GPU local) | H100/H200 class | Open |
| Qwen3.5 122B FP16 | Estimated | 150GB | 162GB | 128GB+ | Cloud / larger GPU | Cloud-first (no practical single-GPU local) | H100/H200 class | Open |
| Qwen3.5 122B Q4 | Estimated | 70GB | 80GB | 128GB+ | Cloud / larger GPU | Dual RTX 4090 (model parallel) | A100 80GB | Open |
| Qwen3.5 122B Q5 | Estimated | 80GB | 82GB | 128GB+ | Cloud / larger GPU | Cloud-first (no practical single-GPU local) | H100/H200 class | Open |
| Qwen3.5 122B Q8 | Estimated | 144GB | 154GB | 128GB+ | Cloud / larger GPU | Cloud-first (no practical single-GPU local) | H100/H200 class | Open |
| GPT-OSS 120B CLOUD | Estimated | 70GB | 78GB | 128GB+ | Cloud / larger GPU | Dual RTX 4090 (model parallel) | A100 80GB | Open |
| GPT-OSS 120B FP16 | Estimated | 80GB | 92GB | 128GB+ | Cloud / larger GPU | Cloud-first (no practical single-GPU local) | H100/H200 class | Open |
| GPT-OSS 120B Q4 | Estimated | 70GB | 80GB | 128GB+ | Cloud / larger GPU | Dual RTX 4090 (model parallel) | A100 80GB | Open |
| GPT-OSS 120B Q5 | Estimated | 80GB | 82GB | 128GB+ | Cloud / larger GPU | Cloud-first (no practical single-GPU local) | H100/H200 class | Open |
| GPT-OSS 120B Q8 | Estimated | 74GB | 84GB | 128GB+ | Cloud / larger GPU | Cloud-first (no practical single-GPU local) | H100/H200 class | Open |
| DeepSeek-R1 70B FP16 | Estimated | 50GB | 62GB | 128GB+ | Cloud / larger GPU | Dual RTX 4090 (model parallel) | A100 80GB | Open |
| DeepSeek-R1 70B Q4 | Estimated | 38GB | 48GB | 64GB+ | Cloud / larger GPU | RTX 6000 Ada 48GB | A100 80GB | Open |
| DeepSeek-R1 70B Q5 | Estimated | 40GB | 50GB | 128GB+ | Cloud / larger GPU | Dual RTX 4090 (model parallel) | A100 80GB | Open |
| DeepSeek-R1 70B Q8 | Estimated | 44GB | 54GB | 128GB+ | Cloud / larger GPU | Dual RTX 4090 (model parallel) | A100 80GB | Open |
| DeepSeek-R1 67B FP16 | Estimated | 50GB | 62GB | 128GB+ | Cloud / larger GPU | Dual RTX 4090 (model parallel) | A100 80GB | Open |
| DeepSeek-R1 67B Q4 | Estimated | 38GB | 48GB | 64GB+ | Cloud / larger GPU | RTX 6000 Ada 48GB | A100 80GB | Open |
| DeepSeek-R1 67B Q5 | Estimated | 40GB | 50GB | 128GB+ | Cloud / larger GPU | Dual RTX 4090 (model parallel) | A100 80GB | Open |
| DeepSeek-R1 67B Q8 | Estimated | 44GB | 54GB | 128GB+ | Cloud / larger GPU | Dual RTX 4090 (model parallel) | A100 80GB | Open |
| DeepSeek-V3 67B FP16 | Estimated | 50GB | 62GB | 128GB+ | Cloud / larger GPU | Dual RTX 4090 (model parallel) | A100 80GB | Open |
| DeepSeek-V3 67B Q4 | Estimated | 38GB | 48GB | 64GB+ | Cloud / larger GPU | RTX 6000 Ada 48GB | A100 80GB | Open |
| DeepSeek-V3 67B Q5 | Estimated | 40GB | 50GB | 128GB+ | Cloud / larger GPU | Dual RTX 4090 (model parallel) | A100 80GB | Open |
| DeepSeek-V3 67B Q8 | Estimated | 44GB | 54GB | 128GB+ | Cloud / larger GPU | Dual RTX 4090 (model parallel) | A100 80GB | Open |
| Qwen3.5 35B FP16 | Estimated | 50GB | 62GB | 128GB+ | Cloud / larger GPU | Dual RTX 4090 (model parallel) | A100 80GB | Open |
| Qwen3.5 35B Q4 | Estimated | 22GB | 24GB | 32GB+ | Comfortable | RTX 3090 24GB | A6000 48GB | Open |
| Qwen3.5 35B Q5 | Estimated | 24GB | 26GB | 64GB+ | Borderline | RTX 6000 Ada 48GB | A100 80GB | Open |
| Qwen3.5 35B Q8 | Estimated | 44GB | 54GB | 128GB+ | Cloud / larger GPU | Dual RTX 4090 (model parallel) | A100 80GB | Open |
| Qwen3.6 35B FP16 | Estimated | 50GB | 62GB | 128GB+ | Cloud / larger GPU | Dual RTX 4090 (model parallel) | A100 80GB | Open |
| Qwen3.6 35B Q4 | Estimated | 22GB | 24GB | 32GB+ | Comfortable | RTX 3090 24GB | A6000 48GB | Open |
| Qwen3.6 35B Q5 | Estimated | 24GB | 26GB | 64GB+ | Borderline | RTX 6000 Ada 48GB | A100 80GB | Open |
| Qwen3.6 35B Q8 | Estimated | 44GB | 54GB | 128GB+ | Cloud / larger GPU | Dual RTX 4090 (model parallel) | A100 80GB | Open |
| DeepSeek-R1 32B FP16 | Estimated | 30GB | 42GB | 64GB+ | Cloud / larger GPU | RTX 6000 Ada 48GB | A100 80GB | Open |
| DeepSeek-R1 32B Q4 | Measured | 20GB | 24GB | 32GB+ | Comfortable | RTX 3090 24GB | A6000 48GB | Open |
| DeepSeek-R1 32B Q5 | Estimated | 24GB | 26GB | 64GB+ | Borderline | RTX 6000 Ada 48GB | A100 80GB | Open |
| DeepSeek-R1 32B Q8 | Estimated | 24GB | 34GB | 64GB+ | Borderline | RTX 6000 Ada 48GB | A100 80GB | Open |
| QwQ 32B FP16 | Estimated | 30GB | 42GB | 64GB+ | Cloud / larger GPU | RTX 6000 Ada 48GB | A100 80GB | Open |
| QwQ 32B Q4 | Estimated | 18GB | 28GB | 64GB+ | Borderline | RTX 6000 Ada 48GB | A100 80GB | Open |
| QwQ 32B Q5 | Estimated | 20GB | 30GB | 64GB+ | Borderline | RTX 6000 Ada 48GB | A100 80GB | Open |
| QwQ 32B Q8 | Estimated | 24GB | 34GB | 64GB+ | Borderline | RTX 6000 Ada 48GB | A100 80GB | Open |
| Magistral 24B FP16 | Estimated | 30GB | 42GB | 64GB+ | Cloud / larger GPU | RTX 6000 Ada 48GB | A100 80GB | Open |
| Magistral 24B Q4 | Estimated | 18GB | 22GB | 32GB+ | Comfortable | RTX 3090 24GB | A6000 48GB | Open |
| Magistral 24B Q5 | Estimated | 22GB | 24GB | 32GB+ | Comfortable | RTX 3090 24GB | A6000 48GB | Open |
| Magistral 24B Q8 | Estimated | 24GB | 34GB | 64GB+ | Borderline | RTX 6000 Ada 48GB | A100 80GB | Open |
| GPT-OSS 20B CLOUD | Estimated | 20GB | 28GB | 64GB+ | Borderline | RTX 6000 Ada 48GB | A100 80GB | Open |
| GPT-OSS 20B FP16 | Estimated | 30GB | 42GB | 64GB+ | Cloud / larger GPU | RTX 6000 Ada 48GB | A100 80GB | Open |
| GPT-OSS 20B Q4 | Estimated | 16GB | 20GB | 32GB+ | Comfortable | RTX 3090 24GB | A6000 48GB | Open |
| GPT-OSS 20B Q5 | Estimated | 20GB | 22GB | 32GB+ | Comfortable | RTX 3090 24GB | A6000 48GB | Open |
| GPT-OSS 20B Q8 | Estimated | 24GB | 34GB | 64GB+ | Borderline | RTX 6000 Ada 48GB | A100 80GB | Open |
| DeepSeek-R1 14B FP16 | Estimated | 22GB | 34GB | 64GB+ | Borderline | RTX 6000 Ada 48GB | A100 80GB | Open |
| DeepSeek-R1 14B Q4 | Estimated | 10GB | 20GB | 32GB+ | Comfortable | RTX 3090 24GB | A6000 48GB | Open |
| DeepSeek-R1 14B Q5 | Estimated | 12GB | 22GB | 32GB+ | Comfortable | RTX 3090 24GB | A6000 48GB | Open |
| DeepSeek-R1 14B Q8 | Estimated | 16GB | 26GB | 64GB+ | Borderline | RTX 6000 Ada 48GB | A100 80GB | Open |
| Phi-3 14B FP16 | Estimated | 22GB | 34GB | 64GB+ | Borderline | RTX 6000 Ada 48GB | A100 80GB | Open |
| Phi-3 14B Q4 | Estimated | 10GB | 20GB | 32GB+ | Comfortable | RTX 3090 24GB | A6000 48GB | Open |
| Phi-3 14B Q5 | Estimated | 12GB | 22GB | 32GB+ | Comfortable | RTX 3090 24GB | A6000 48GB | Open |
| Phi-3 14B Q8 | Estimated | 16GB | 26GB | 64GB+ | Borderline | RTX 6000 Ada 48GB | A100 80GB | Open |
| Phi-4 14B FP16 | Estimated | 22GB | 34GB | 64GB+ | Borderline | RTX 6000 Ada 48GB | A100 80GB | Open |
| Phi-4 14B Q4 | Estimated | 10GB | 20GB | 32GB+ | Comfortable | RTX 3090 24GB | A6000 48GB | Open |
| Phi-4 14B Q5 | Estimated | 12GB | 22GB | 32GB+ | Comfortable | RTX 3090 24GB | A6000 48GB | Open |
| Phi-4 14B Q8 | Estimated | 16GB | 26GB | 64GB+ | Borderline | RTX 6000 Ada 48GB | A100 80GB | Open |
| Phi-4 Reasoning 14B FP16 | Estimated | 22GB | 34GB | 64GB+ | Borderline | RTX 6000 Ada 48GB | A100 80GB | Open |
| Phi-4 Reasoning 14B Q4 | Estimated | 10GB | 20GB | 32GB+ | Comfortable | RTX 3090 24GB | A6000 48GB | Open |
| Phi-4 Reasoning 14B Q5 | Estimated | 12GB | 22GB | 32GB+ | Comfortable | RTX 3090 24GB | A6000 48GB | Open |
| Phi-4 Reasoning 14B Q8 | Estimated | 16GB | 26GB | 64GB+ | Borderline | RTX 6000 Ada 48GB | A100 80GB | Open |
| OLMo 2 13B FP16 | Estimated | 22GB | 34GB | 64GB+ | Borderline | RTX 6000 Ada 48GB | A100 80GB | Open |
| OLMo 2 13B Q4 | Estimated | 10GB | 20GB | 32GB+ | Comfortable | RTX 3090 24GB | A6000 48GB | Open |
| OLMo 2 13B Q5 | Estimated | 12GB | 22GB | 32GB+ | Comfortable | RTX 3090 24GB | A6000 48GB | Open |
| OLMo 2 13B Q8 | Estimated | 16GB | 26GB | 64GB+ | Borderline | RTX 6000 Ada 48GB | A100 80GB | Open |
| DeepSeek-R1 8B FP16 | Estimated | 18GB | 30GB | 64GB+ | Borderline | RTX 6000 Ada 48GB | A100 80GB | Open |
| DeepSeek-R1 8B Q4 | Estimated | 6GB | 16GB | 32GB+ | Comfortable | RTX 3090 24GB | A6000 48GB | Open |
| DeepSeek-R1 8B Q5 | Estimated | 8GB | 18GB | 32GB+ | Comfortable | RTX 3090 24GB | A6000 48GB | Open |
| DeepSeek-R1 8B Q8 | Estimated | 12GB | 22GB | 32GB+ | Comfortable | RTX 3090 24GB | A6000 48GB | Open |
| DeepSeek-R1 7B FP16 | Estimated | 18GB | 30GB | 64GB+ | Borderline | RTX 6000 Ada 48GB | A100 80GB | Open |
| DeepSeek-R1 7B Q4 | Estimated | 6GB | 16GB | 32GB+ | Comfortable | RTX 3090 24GB | A6000 48GB | Open |
| DeepSeek-R1 7B Q5 | Estimated | 8GB | 18GB | 32GB+ | Comfortable | RTX 3090 24GB | A6000 48GB | Open |
| DeepSeek-R1 7B Q8 | Estimated | 12GB | 22GB | 32GB+ | Comfortable | RTX 3090 24GB | A6000 48GB | Open |
| OLMo 2 7B FP16 | Estimated | 18GB | 30GB | 64GB+ | Borderline | RTX 6000 Ada 48GB | A100 80GB | Open |
| OLMo 2 7B Q4 | Estimated | 6GB | 16GB | 32GB+ | Comfortable | RTX 3090 24GB | A6000 48GB | Open |
| OLMo 2 7B Q5 | Estimated | 8GB | 18GB | 32GB+ | Comfortable | RTX 3090 24GB | A6000 48GB | Open |
| OLMo 2 7B Q8 | Estimated | 12GB | 22GB | 32GB+ | Comfortable | RTX 3090 24GB | A6000 48GB | Open |
| Phi-3 3.8B FP16 | Estimated | 16GB | 28GB | 64GB+ | Borderline | RTX 6000 Ada 48GB | A100 80GB | Open |
| Phi-3 3.8B Q4 | Estimated | 4GB | 14GB | 32GB+ | Comfortable | RTX 3090 24GB | A6000 48GB | Open |
| Phi-3 3.8B Q5 | Estimated | 6GB | 16GB | 32GB+ | Comfortable | RTX 3090 24GB | A6000 48GB | Open |
| Phi-3 3.8B Q8 | Estimated | 10GB | 20GB | 32GB+ | Comfortable | RTX 3090 24GB | A6000 48GB | Open |
| Granite 3.1 MoE 3B FP16 | Estimated | 16GB | 28GB | 64GB+ | Borderline | RTX 6000 Ada 48GB | A100 80GB | Open |
| Granite 3.1 MoE 3B Q4 | Estimated | 4GB | 14GB | 32GB+ | Comfortable | RTX 3090 24GB | A6000 48GB | Open |
| Granite 3.1 MoE 3B Q5 | Estimated | 6GB | 16GB | 32GB+ | Comfortable | RTX 3090 24GB | A6000 48GB | Open |
| Granite 3.1 MoE 3B Q8 | Estimated | 10GB | 20GB | 32GB+ | Comfortable | RTX 3090 24GB | A6000 48GB | Open |
| DeepSeek-R1 1.5B FP16 | Estimated | 14GB | 26GB | 64GB+ | Borderline | RTX 6000 Ada 48GB | A100 80GB | Open |
| DeepSeek-R1 1.5B Q4 | Estimated | 2GB | 12GB | 16GB+ | Comfortable | RTX 3090 24GB | A6000 48GB | Open |
| DeepSeek-R1 1.5B Q5 | Estimated | 4GB | 14GB | 32GB+ | Comfortable | RTX 3090 24GB | A6000 48GB | Open |
| DeepSeek-R1 1.5B Q8 | Estimated | 8GB | 18GB | 32GB+ | Comfortable | RTX 3090 24GB | A6000 48GB | Open |
| Granite 3.1 MoE 1B FP16 | Estimated | 14GB | 26GB | 64GB+ | Borderline | RTX 6000 Ada 48GB | A100 80GB | Open |
| Granite 3.1 MoE 1B Q4 | Estimated | 2GB | 12GB | 16GB+ | Comfortable | RTX 3090 24GB | A6000 48GB | Open |
| Granite 3.1 MoE 1B Q5 | Estimated | 4GB | 14GB | 32GB+ | Comfortable | RTX 3090 24GB | A6000 48GB | Open |
| Granite 3.1 MoE 1B Q8 | Estimated | 8GB | 18GB | 32GB+ | Comfortable | RTX 3090 24GB | A6000 48GB | Open |
We may earn a commission if you click links on this page.