What Can I Run AI?

DeepSeek-R1-Distill-Qwen-32B vs Qwen3-32B: VRAM and speed for running locally

Side by side: how much memory each model needs at Q4, Q8 and FP16, how the KV cache grows with context, the cheapest GPU and Mac that run each, and estimated speed on 11 common machines.

Qwen3-32B is lighter: about 23.8 GB at Q4 against 23.8 GB for DeepSeek-R1-Distill-Qwen-32B, though both run on 16 of the devices we track. It should also be faster on the same hardware, since it reads 32.8B parameters per token against 32.8B. Which one gives better answers depends on your task, so check benchmarks for that; this page covers what it takes to run them.

DeepSeek-R1-Distill-Qwen-32B vs Qwen3-32B: specs and memory
DeepSeek-R1-Distill-Qwen-32BQwen3-32B
Makerdeepseek-aiQwen
Licensemitapache-2.0
TypeDenseDense
Total parameters32.8B32.8B
Active per token32.8B all32.8B all
Memory at Q4_K_M23.8 GB23.8 GB
Memory at Q8_041.0 GB41.0 GB
Memory at FP1676.4 GB76.4 GB
KV cache growth262 MB per 1k tokens, FP16262 MB per 1k tokens, FP16
Memory at Q4, 32k context31.4 GB 8.6 GB cache31.4 GB 8.6 GB cache
Cheapest GPU at Q4Intel Arc Pro B70 32GB $949Intel Arc Pro B70 32GB $949
Cheapest Mac / mini PC at Q4Mac mini M5 Pro 64GB $3,199Mac mini M5 Pro 64GB $3,199
Cheapest GPU at Q8NVIDIA RTX 6000 Ada 48GB $6,799NVIDIA RTX 6000 Ada 48GB $6,799
Tracked devices that run it at Q416 of 8216 of 82

Estimated speed on the same hardware

Decode tokens per second at Q4_K_M from memory bandwidth (how we estimate). MoE figures are upper bounds ("up to"). Estimates, not benchmarks.

Estimated tokens per second, DeepSeek-R1-Distill-Qwen-32B vs Qwen3-32B
DeviceDeepSeek-R1-Distill-Qwen-32BQwen3-32B
GeForce RTX 3060 12GB12 GB usable · 360 GB/s · $329doesn’t fitdoesn’t fit
GeForce RTX 4060 Ti 16GB16 GB usable · 288 GB/s · $499doesn’t fitdoesn’t fit
GeForce RTX 508016 GB usable · 960 GB/s · $999doesn’t fitdoesn’t fit
GeForce RTX 309024 GB usable · 936 GB/s · $1,499tight fittight fit
Radeon RX 7900 XTX24 GB usable · 960 GB/s · $999tight fittight fit
GeForce RTX 509032 GB usable · 1792 GB/s · $1,99954–7754–77
Mac mini M6 24GB18 GB usable · 170 GB/s · $1,299doesn’t fitdoesn’t fit
Mac mini M6 32GB24 GB usable · 170 GB/s · $1,799tight fittight fit
Mac mini M5 Pro 64GB48 GB usable · 307 GB/s · $3,1997–127–12
Framework Desktop (Ryzen AI Max+ 395) 128GB96 GB usable · 256 GB/s · $3,4496–106–10
MacBook Pro M5 Max 128GB96 GB usable · 614 GB/s · $6,69914–2514–25

Quick answers

Which needs less VRAM, DeepSeek-R1-Distill-Qwen-32B or Qwen3-32B?
Qwen3-32B: about 23.8 GB at Q4_K_M, against 23.8 GB for DeepSeek-R1-Distill-Qwen-32B. At Q8_0 it's 41.0 GB vs 41.0 GB.
Which is faster locally, DeepSeek-R1-Distill-Qwen-32B or Qwen3-32B?
Decode speed is limited by how many bytes are read per token. Qwen3-32B reads 32.8B parameters per token and DeepSeek-R1-Distill-Qwen-32B reads 32.8B, so Qwen3-32B should be faster on the same machine. Our estimate on the GeForce RTX 5090 at Q4 is 54–77 tok/s vs 54–77 tok/s. These are estimates, not benchmarks.
What is the cheapest GPU for DeepSeek-R1-Distill-Qwen-32B and Qwen3-32B?
At Q4: Intel Arc Pro B70 32GB ($949) for DeepSeek-R1-Distill-Qwen-32B; Intel Arc Pro B70 32GB ($949) for Qwen3-32B. Prices are dated list-price snapshots.
Can a Mac run DeepSeek-R1-Distill-Qwen-32B and Qwen3-32B?
The cheapest Mac or mini PC we track for DeepSeek-R1-Distill-Qwen-32B is the Mac mini M5 Pro 64GB ($3,199). For Qwen3-32B it's the Mac mini M5 Pro 64GB ($3,199).
Which is better, DeepSeek-R1-Distill-Qwen-32B or Qwen3-32B?
We don't score answer quality. This comparison covers memory, context cache, hardware and speed. Pick the one that fits your hardware, then test both on your own prompts.

Related comparisons

All comparisons · MoE vs dense · Q4 vs Q8 vs FP16 · Context length and VRAM · VRAM calculator

New models, and what they need to run

One short email a week: fresh open models with their VRAM numbers, plus hardware worth watching.

Or subscribe here