Qwen · 32B to 80B
Qwen2.5 32B Instruct
32.76 billion parameters across 64 layers. At four-bit precision the weights alone come to 18.42 GB, before any conversation is loaded. Grouped-query attention keeps the cache small, so long contexts cost less here than on models of the same size.
Memory needed, by quantisation
At an 8k context window. Quality is our rough ranking of how much the compression costs you: anything at or above Q5 is hard to tell apart from the original in normal use.
| Quantisation | File size | KV cache | Total VRAM | Quality | What it costs you |
|---|---|---|---|---|---|
| Q8_0 | 32.42 GB | 2 GB | 35.33 GB | 99% | Lossless in practice. Use it when the memory is there. |
| Q6_K | 25.02 GB | 2 GB | 27.93 GB | 98% | Very close to Q8 for two thirds of the size. |
| Q5_K_M | 21.7 GB | 2 GB | 24.61 GB | 96% | The quality-first choice when Q6 will not fit. |
| Q4_K_M | 18.42 GB | 2 GB | 21.33 GB | 93% | The default. Best size-to-quality ratio for local use. |
| IQ4_XS | 16.21 GB | 2 GB | 19.12 GB | 91% | Importance-matrix 4-bit. Q4_K_S quality, smaller file. |
| Q3_K_M | 14.91 GB | 2 GB | 17.83 GB | 86% | Degradation starts to show. A way to fit one size up. |
| IQ3_XXS | 11.67 GB | 2 GB | 14.58 GB | 79% | Aggressive. Only worth it on very large models. |
What a longer conversation costs
Same model at Q4_K_M, only the context window changes. This model uses sliding-window attention, so most layers stop growing past 128k tokens and the bill flattens out.
| Context | KV cache | Total VRAM | Fits in 8 GB | Fits in 12 GB | Fits in 24 GB |
|---|---|---|---|---|---|
| 4k | 1 GB | 20.18 GB | no | no | yes |
| 8k | 2 GB | 21.33 GB | no | no | yes |
| 16k | 4 GB | 23.65 GB | no | no | no |
| 32k | 8 GB | 28.27 GB | no | no | no |
The "fits" columns allow for the roughly 0.8 GB Windows keeps for the desktop.
Which hardware runs Qwen2.5 32B Instruct
50 of 118 consumer devices run it at a quantisation worth using, at an 8k context window. Another 44 can load it only by compressing the weights far enough to damage the model, marked with a warning below.
| Device | Memory | Verdict | Quantisation | Used | Speed |
|---|---|---|---|---|---|
| RTX PRO 6000 Blackwell NVIDIA | 96 GB | Runs great | Q8_0 | 35.33 GB | 42.7 t/s |
| RTX 6000 Ada Generation NVIDIA | 48 GB | Runs well | Q8_0 | 35.33 GB | 22.9 t/s |
| GeForce RTX 5090 NVIDIA | 32 GB | Runs great | Q6_K | 27.93 GB | 54.4 t/s |
| Apple M3 Ultra 96GB Apple | 96 GB | Runs well | Q8_0 | 35.03 GB | 18.6 t/s |
| Apple M3 Ultra 256GB Apple | 256 GB | Runs well | Q8_0 | 35.03 GB | 18.6 t/s |
| Apple M3 Ultra 512GB Apple | 512 GB | Runs well | Q8_0 | 35.03 GB | 18.6 t/s |
| RTX A6000 NVIDIA | 48 GB | Runs well | Q8_0 | 35.33 GB | 18.3 t/s |
| Radeon PRO W7900 AMD | 48 GB | Runs well | Q8_0 | 35.33 GB | 18.1 t/s |
| Apple M1 Ultra 64GB Apple | 64 GB | Runs well | Q8_0 | 35.03 GB | 18.1 t/s |
| Apple M1 Ultra 128GB Apple | 128 GB | Runs well | Q8_0 | 35.03 GB | 18.1 t/s |
| Apple M2 Ultra 64GB Apple | 64 GB | Runs well | Q8_0 | 35.03 GB | 18.1 t/s |
| Apple M2 Ultra 128GB Apple | 128 GB | Runs well | Q8_0 | 35.03 GB | 18.1 t/s |
| Apple M2 Ultra 192GB Apple | 192 GB | Runs well | Q8_0 | 35.03 GB | 18.1 t/s |
| Apple M4 Max 36GB Apple | 36 GB | Just fits | Q5_K_M | 24.31 GB | 18 t/s |
| GeForce RTX 4090 NVIDIA | 24 GB | Runs great | Q4_K_M | 21.33 GB | 40.5 t/s |
| GeForce RTX 3090 Ti NVIDIA | 24 GB | Runs great | Q4_K_M | 21.33 GB | 40.5 t/s |
| GeForce RTX 3090 NVIDIA | 24 GB | Runs great | Q4_K_M | 21.33 GB | 37.6 t/s |
| GeForce RTX 5090 Laptop NVIDIA | 24 GB | Runs great | Q4_K_M | 21.33 GB | 36 t/s |
| Radeon RX 7900 XTX AMD | 24 GB | Runs great | Q4_K_M | 21.33 GB | 33.8 t/s |
| RTX A5000 NVIDIA | 24 GB | Runs great | Q4_K_M | 21.33 GB | 30.8 t/s |
| Radeon RX 7900 XT AMD | 20 GB | Just fits | IQ4_XS | 19.12 GB | 31.6 t/s |
| Apple M4 Max 48GB Apple | 48 GB | Just fits | Q8_0 | 35.03 GB | 12.4 t/s |
| Apple M4 Max 64GB Apple | 64 GB | Runs well | Q8_0 | 35.03 GB | 12.4 t/s |
| Apple M4 Max 128GB Apple | 128 GB | Runs well | Q8_0 | 35.03 GB | 12.4 t/s |
| Apple M3 Max 36GB Apple | 36 GB | Just fits | Q5_K_M | 24.31 GB | 13.2 t/s |
| Apple M1 Max 32GB Apple | 32 GB | Runs well | Q4_K_M | 21.03 GB | 15.3 t/s |
| Apple M2 Max 32GB Apple | 32 GB | Runs well | Q4_K_M | 21.03 GB | 15.3 t/s |
| Apple M1 Max 64GB Apple | 64 GB | Runs well | Q6_K | 27.63 GB | 11.5 t/s |
| Apple M2 Max 64GB Apple | 64 GB | Runs well | Q6_K | 27.63 GB | 11.5 t/s |
| Apple M2 Max 96GB Apple | 96 GB | Runs well | Q6_K | 27.63 GB | 11.5 t/s |
| Apple M3 Max 48GB Apple | 48 GB | Runs well | Q6_K | 27.63 GB | 11.5 t/s |
| Apple M3 Max 64GB Apple | 64 GB | Runs well | Q6_K | 27.63 GB | 11.5 t/s |
| Apple M3 Max 96GB Apple | 96 GB | Runs well | Q6_K | 27.63 GB | 11.5 t/s |
| Apple M3 Max 128GB Apple | 128 GB | Runs well | Q6_K | 27.63 GB | 11.5 t/s |
| NVIDIA DGX Spark 128GB NVIDIA | 128 GB | Runs well | Q4_K_M | 21.33 GB | 11 t/s |
| Apple M4 Pro 48GB Apple | 48 GB | Runs well | Q4_K_M | 21.03 GB | 10.4 t/s |
| Apple M4 Pro 64GB Apple | 64 GB | Runs well | Q4_K_M | 21.03 GB | 10.4 t/s |
| GeForce RTX 5080 NVIDIA | 16 GB | Just fits | IQ3_XXS ! | 14.58 GB | 57.6 t/s |
| GeForce RTX 5070 Ti NVIDIA | 16 GB | Just fits | IQ3_XXS ! | 14.58 GB | 53.7 t/s |
| GeForce RTX 5080 Laptop NVIDIA | 16 GB | Just fits | IQ3_XXS ! | 14.58 GB | 46.1 t/s |
| GeForce RTX 4080 SUPER NVIDIA | 16 GB | Just fits | IQ3_XXS ! | 14.58 GB | 44.1 t/s |
| GeForce RTX 4080 NVIDIA | 16 GB | Just fits | IQ3_XXS ! | 14.58 GB | 43 t/s |
| GeForce RTX 4070 Ti SUPER NVIDIA | 16 GB | Just fits | IQ3_XXS ! | 14.58 GB | 40.3 t/s |
| GeForce RTX 4090 Laptop NVIDIA | 16 GB | Just fits | IQ3_XXS ! | 14.58 GB | 34.6 t/s |
| Radeon RX 9070 XT AMD | 16 GB | Just fits | IQ3_XXS ! | 14.58 GB | 34 t/s |
| Radeon RX 9070 AMD | 16 GB | Just fits | IQ3_XXS ! | 14.58 GB | 34 t/s |
| Radeon RX 7800 XT AMD | 16 GB | Just fits | IQ3_XXS ! | 14.58 GB | 32.9 t/s |
| Radeon RX 7900 GRE AMD | 16 GB | Just fits | IQ3_XXS ! | 14.58 GB | 30.3 t/s |
| Radeon RX 6900 XT AMD | 16 GB | Just fits | IQ3_XXS ! | 14.58 GB | 27 t/s |
| Radeon RX 6800 AMD | 16 GB | Just fits | IQ3_XXS ! | 14.58 GB | 27 t/s |
| GeForce RTX 5060 Ti 16GB NVIDIA | 16 GB | Just fits | IQ3_XXS ! | 14.58 GB | 26.9 t/s |
| RTX A4000 NVIDIA | 16 GB | Just fits | IQ3_XXS ! | 14.58 GB | 26.9 t/s |
| Arc A770 16GB Intel | 16 GB | Just fits | IQ3_XXS ! | 14.58 GB | 26.6 t/s |
| Jetson AGX Orin 32GB NVIDIA | 32 GB | Fits, but slow | Q4_K_M | 21.33 GB | 8.2 t/s |
| Ryzen AI Max+ 395 32GB AMD | 32 GB | Fits, but slow | Q4_K_M | 21.33 GB | 8.1 t/s |
| GeForce RTX 4060 Ti 16GB NVIDIA | 16 GB | Just fits | IQ3_XXS ! | 14.58 GB | 17.3 t/s |
| Apple M1 Pro 32GB Apple | 32 GB | Fits, but slow | Q4_K_M | 21.03 GB | 7.6 t/s |
| Apple M2 Pro 32GB Apple | 32 GB | Fits, but slow | Q4_K_M | 21.03 GB | 7.6 t/s |
| Apple M4 Pro 24GB Apple | 24 GB | Runs well | IQ3_XXS ! | 14.28 GB | 15.6 t/s |
| Radeon RX 7600 XT AMD | 16 GB | Just fits | IQ3_XXS ! | 14.58 GB | 15.2 t/s |
What to buy to run Qwen2.5 32B Instruct
The cheapest hardware that runs it at a quantisation worth using and a speed you would not resent, at an 8k context window.
Cheapest that works
Radeon RX 7900 XT
$650
IQ4_XS · uses 19.12 GB of its 20 GB · about 31.6 tokens/s
Best value
GeForce RTX 3090
$700 used
Q4_K_M · about 37.6 tokens/s · 936 GB/s
| Device | Price | Memory | Runs it at | Speed | |
|---|---|---|---|---|---|
| Radeon RX 7900 XT AMD | $650 | 20 GB | IQ4_XS | 31.6 t/s | Check price |
| GeForce RTX 3090 NVIDIA | $700 used | 24 GB | Q4_K_M | 37.6 t/s | Check price |
| GeForce RTX 3090 Ti NVIDIA | $800 used | 24 GB | Q4_K_M | 40.5 t/s | Check price |
| Apple M1 Max 32GB Apple · whole machine | $1,400 | 32 GB | Q4_K_M | 15.3 t/s | Check price |
| Apple M1 Max 64GB Apple · whole machine | $1,900 | 64 GB | Q4_K_M | 15.3 t/s | Check price |
| GeForce RTX 5090 NVIDIA | $2,200 | 32 GB | Q6_K | 54.4 t/s | Check price |
| Apple M1 Ultra 128GB Apple · whole machine | $3,400 | 128 GB | Q8_0 | 18.1 t/s | Check price |
| Apple M3 Ultra 256GB Apple · whole machine | $5,599 | 256 GB | Q8_0 | 18.6 t/s | Check price |
Indicative prices reviewed 2026-09-01; used prices are marketplace typical. Options that cost more than a cheaper one with no more memory and no more speed are hidden. Price links are Amazon affiliate links. Change the standard, the context window or the budget in the buying tool.
These numbers also cover
Fine-tunes share their base model's architecture, so they need exactly the same memory. If you are looking for one of these, the figures above apply unchanged.
deepseek-ai/DeepSeek-R1-Distill-Qwen-32BQwen/Qwen2.5-Coder-32B-InstructQwen/Qwen2.5-32B
Where it sits in the catalogue
Source: Qwen/Qwen2.5-32B-Instruct on Hugging Face. Downloaded 2.0M times in the last month. Published 2024-09-17. Architecture figures are read from the repository's own configuration file, so they move when the model does.
Qwen3 32B
32.76B · 40k context
Qwen3 30B A3B
30.53B (3.34B active) · 40k context
Qwen3.5 35B A3B
35.95B (2.9B active) · 256k context
Qwen3.5 27B
27.78B · 256k context
Qwen3 14B
14.77B · 40k context
Qwen2.5 14B Instruct
14.77B · 32k context
Qwen1.5 MoE A2.7B
14.32B (2.69B active) · 8k context
Qwen3.5 9B
9.65B · 256k context