The NVIDIA RTX 5090 D is a consumer-grade GPU built on the Blackwell architecture, featuring 32 GB of memory with 1,792 GB/s of memory bandwidth and a 575 W TDP. It delivers 104.8 TFLOPS of FP32 performance.
Hardware specifications
Same across all providers
Median price across all providers
0 offerings from 0 providers
Sorted by price ascending. Compare configurations side-by-side.
| Provider | Count | vCPU | RAM | Region | Per GPU hour | Total/hr | Action |
|---|---|---|---|---|---|---|---|
| Try broadening your filters or clear all filters to start over. | |||||||
LLMs that fit on the NVIDIA RTX 5090 D
Estimated VRAM for popular open-weight LLMs at 16-bit (FP16/BF16), 8-bit (FP8/INT8) and 4-bit (INT4/MXFP4) precision, against 32 GB per GPU. Showing the 12 largest that fit.
| Model | Parameters | 16-bit | 8-bit | 4-bit |
|---|---|---|---|---|
| Qwen3.6 35B-A3B Alibaba | 35B 3B active | 86 GB Too large | 47 GB Too large | 28 GB 1× · 4.4 GB free |
| Qwen3 32B Alibaba | 32.8B | 79 GB Too large | 43 GB Too large | 25 GB 1× · 6.8 GB free |
| Gemma 4 31B Google | 31.3B | 75 GB Too large | 41 GB Too large | 24 GB 1× · 8 GB free |
| GLM-4.7 Flash Z.ai | 31.2B | 75 GB Too large | 41 GB Too large | 24 GB 1× · 8 GB free |
| Qwen3.6 27B Alibaba | 27.8B | 67 GB Too large | 37 GB Tight on 1× · 4% spare | 21 GB 1× · 11 GB free |
| Mistral Small 3.2 24B Mistral AI | 24B | 58 GB Too large | 32 GB Barely fits on 1× | 18 GB 1× · 14 GB free |
| gpt-oss-20b OpenAI | 21B 3.6B active | — Not released | — Not released | 17 GB 1× · 15 GB free |
| Gemma 3 12B Google | 12.2B | 29 GB 1× · 2.8 GB free | 16 GB 1× · 16 GB free | 9.4 GB 1× · 23 GB free |
| Granite 4.1 8B IBM | 8.8B | 21 GB 1× · 11 GB free | 12 GB 1× · 20 GB free | 6.8 GB 1× · 25 GB free |
| Qwen3 8B Alibaba | 8.2B | 20 GB 1× · 12 GB free | 11 GB 1× · 21 GB free | 6.3 GB 1× · 26 GB free |
| Llama 3.1 8B Instruct Meta | 8B | 19 GB 1× · 13 GB free | 11 GB 1× · 21 GB free | 6.2 GB 1× · 26 GB free |
| Mistral 7B Instruct v0.3 Mistral AI | 7.2B | 17 GB 1× · 15 GB free | 9.6 GB 1× · 22 GB free | 5.6 GB 1× · 26 GB free |
Estimates: model weights plus 20% for KV cache and runtime overhead, at short context lengths. Long contexts and large batches need more. Rows marked “tight” hold the weights but not that full margin. A ×N figure is the total VRAM across N of these GPUs and makes no claim about interconnect throughput. How we estimate this · Start from a model instead