Can the RTX 5090 Run Qwen3 32B?
Runs
Yes. Qwen3 32B at Q4_K_M is a 19.8 GB download (Hugging Face file listing for unsloth/Qwen3-32B-GGUF), and the RTX 5090 has 32 GB of VRAM, leaving about 12.2 GB for the KV cache and runtime. Published runs put generation at a median of 59.7 tokens per second across 2 cited measurements. Fastest cited run: Hardware Corner.
Buy the RTX 5090
GIGABYTE GeForce RTX 5090 WINDFORCE OC 32G Graphics Card, WINDFORCE Cooling System, 32GB…Tracked at $5,500 on 2026-09-08 — price may vary.
Published runs for Qwen3 32B on the RTX 5090
| Generation | Quant | Runtime | Context | Source | Date |
|---|---|---|---|---|---|
| 61.4 tok/s | q4_K_M | llama.cpp | 4K | Hardware Cornerlabelled "qwen3:32b" | 2025-06-01 |
| 58 tok/s | Q4_K_M | llama.cpp | — | DatabaseMartlabelled "Qwen3 32B" | 2025-04-15 |
More about this pair
How much VRAM does Qwen3 32B need?
The Q4_K_M file is 19.8 GB. Add roughly 1.5 GB for the runtime and an ~8K-token context, so about 21.3 GB of VRAM holds it comfortably. Longer contexts grow the KV cache and need more.
What is the fastest Qwen3 32B can generate on the RTX 5090?
Generation reads every weight once per token, so the RTX 5090's 1792 GB/s of memory bandwidth divided by the 19.8 GB file sets a ceiling near 91 tokens per second. Real runs land below it; SpecPicks rejects any published figure above it.
Check another pair
Qwen3 32B on other GPUs
- RTX 4060 — Offload
- RTX 3060 Ti — Offload
- RTX 3060 12GB — Offload
- Intel Arc B580 — Offload
- RTX 4070 SUPER — Offload
- RTX 4060 Ti 16GB — Offload
- RTX 5060 Ti 16GB — Offload
- Intel Arc A770 16GB — Offload
- Radeon RX 9070 XT — Offload
- RTX 4070 Ti SUPER — Offload
- RTX 4080 — Offload
- RTX 5070 Ti — Offload
- RTX 5080 — Offload
- RTX 3090 — Runs, 32.7 tok/s median
- Radeon RX 7900 XTX — Runs, 27 tok/s median
- RTX 4090 — Runs, 38 tok/s median
- RTX A6000 48GB — Runs, 27.5 tok/s median
- RTX PRO 6000 Blackwell 96GB — Runs, 56 tok/s median
Other models on the RTX 5090
- Llama 3.2 1B — Runs
- Llama 3.2 3B — Runs
- Mistral 7B — Runs
- Qwen2.5 7B — Runs
- Llama 3.1 8B — Runs, 150 tok/s median
- DeepSeek-R1-Distill-Llama 8B — Runs
- Qwen3 8B — Runs, 186 tok/s median
- Gemma 2 9B — Runs
- Gemma 3 12B — Runs
- Qwen2.5 14B — Runs, 89.9 tok/s median
- DeepSeek-R1-Distill-Qwen 14B — Runs, 89.1 tok/s median
- Qwen3 14B — Runs, 124 tok/s median
- Phi-4 14B — Runs
- gpt-oss-20b — Runs
- Gemma 3 27B — Runs, 47.3 tok/s median
- Gemma 2 27B — Runs
- Qwen3 30B-A3B — Runs, 234 tok/s median
- Qwen2.5 32B — Runs, 45.1 tok/s median
- DeepSeek-R1-Distill-Qwen 32B — Runs, 49.8 tok/s median
- Llama 3.1 70B — Offload
- Llama 3.3 70B — Offload
- DeepSeek-R1-Distill-Llama 70B — Offload
How the checks work
The fit is the Q4_K_M file size against the card's VRAM, with 1.5 GB kept for the runtime and an ~8K context. A published run counts only if the pair fits, the run is Q4-class, its source does not call it an estimate, it is under the bandwidth ceiling above, and it is not one figure pasted across many cards. The whole matrix is downloadable as a CSV under CC BY 4.0.
As an Amazon Associate, SpecPicks earns from qualifying purchases.
This page is editorial synthesis based on publicly available information. No independent first-party benchmarking is reported.