DeepSeek-R1-Distill-Qwen-32B on NVIDIA H200 141GB SXM (141GB VRAM)
In 4-bit Q4_K_M, DeepSeek-R1-Distill-Qwen-32B uses only 23.14 GB of your 141 GB VRAM (16% utilization) at ~149.7 tokens/sec. In full 16-bit precision, it requires 70.23 GB.
🎛️ Interactive VRAM & KV-Cache Tuner
📊 Real-Time Memory Footprint
23.14 GB / 141 GB🚀 Ready-to-Run Launch Command
ollama run deepseek-r1:32b # Or with llama.cpp server (8K context, Flash Attention): llama-server -hf deepseek-ai/DeepSeek-R1-Distill-Qwen-32B-GGUF -c 8192 -ngl 99 -fa
Need 23.14 GB+ VRAM to run DeepSeek-R1-Distill-Qwen-32B?
Recommended cloud setup for 4-bit: RTX 4090 (24GB) (or 70.23 GB for full FP16). Spin up a pre-configured vLLM / Ollama instance in 60 seconds:
| Provider | Best For | RTX 4090 (24GB) | A100 (80GB) | H100 (80GB) | Bonus / Trial | Action |
|---|---|---|---|---|---|---|
| RunPod | 1-Click vLLM / Ollama Pods | $0.44 / hr | $1.39 / hr | $2.39 / hr | Instant PyTorch / vLLM Pods + Serverless GPU | Deploy on RunPod → |
| Vast.ai | Cheapest Spot GPU Market | $0.29 / hr | $0.95 / hr | $1.79 / hr | Save 50–70% on RTX 4090 / 5090 / A100 rentals | Check Spot GPUs → |
| Vultr Cloud GPU | Enterprise SLA & Bare Metal | — | $1.79 / hr | $2.69 / hr | $300 Free Credit for New Users | Claim $300 Credit → |
| DigitalOcean GPU Droplets | Managed H100 Droplets | — | — | $2.99 / hr | $200 Free Trial (60 Days) | Get $200 Free → |
All Quantization Formats: DeepSeek-R1-Distill-Qwen-32B on NVIDIA H200 141GB SXM
| Quantization Format | Bits/Weight | Weights VRAM | KV Cache (8K) | Total VRAM | Fits on 1x NVIDIA H200 141GB SXM (141GB)? | Est. Speed | GPUs Needed |
|---|---|---|---|---|---|---|---|
| FP16 / BF16 (Unquantized) | 16 bpw | 67.57 GB | 2 GB | 70.23 GB | Yes (50%) | ~45.4 tok/s | 1x NVIDIA H200 141GB SXM |
| FP8 (W8A8 / E4M3) | 8.5 bpw | 35.9 GB | 2 GB | 38.55 GB | Yes (27%) | ~85.4 tok/s | 1x NVIDIA H200 141GB SXM |
| GGUF Q8_0 (8-bit) | 8.5 bpw | 35.9 GB | 2 GB | 38.55 GB | Yes (27%) | ~85.4 tok/s | 1x NVIDIA H200 141GB SXM |
| GGUF Q6_K (6-bit) | 6.56 bpw | 27.7 GB | 2 GB | 30.36 GB | Yes (22%) | ~110.6 tok/s | 1x NVIDIA H200 141GB SXM |
| GGUF Q5_K_M (5-bit) | 5.69 bpw | 24.03 GB | 2 GB | 26.69 GB | Yes (19%) | ~127.6 tok/s | 1x NVIDIA H200 141GB SXM |
| GGUF Q4_K_M / AWQ (4-bit) | 4.85 bpw | 20.48 GB | 2 GB | 23.14 GB | Yes (16%) | ~149.7 tok/s | 1x NVIDIA H200 141GB SXM |
| GGUF IQ4_XS (4.25 bpw) | 4.25 bpw | 17.95 GB | 2 GB | 20.61 GB | Yes (15%) | ~170.8 tok/s | 1x NVIDIA H200 141GB SXM |
| GGUF Q3_K_M (3-bit) | 3.91 bpw | 16.51 GB | 2 GB | 19.17 GB | Yes (14%) | ~185.6 tok/s | 1x NVIDIA H200 141GB SXM |
Frequently Asked Questions (DeepSeek-R1-Distill-Qwen-32B + NVIDIA H200 141GB SXM)
Can NVIDIA H200 141GB SXM (141GB) run DeepSeek-R1-Distill-Qwen-32B locally?
Yes — NVIDIA H200 141GB SXM (141GB) easily runs DeepSeek-R1-Distill-Qwen-32B even in full FP16. In 4-bit Q4_K_M, DeepSeek-R1-Distill-Qwen-32B uses only 23.14 GB of your 141 GB VRAM (16% utilization) at ~149.7 tokens/sec. In full 16-bit precision, it requires 70.23 GB.
How is KV Cache calculated for DeepSeek-R1-Distill-Qwen-32B?
DeepSeek-R1-Distill-Qwen-32B uses Dense-GQA with 64 transformer layers, 40 attention heads, and 8 KV heads (head dimension 128). At 8,192 tokens in FP16, KV cache consumes 2 GB. Enabling 8-bit KV cache (-ctk q8_0 -ctv q8_0 in llama.cpp or --kv-cache-dtype fp8 in vLLM) cuts this in half to 1.00 GB.