unsloth
DeepSeek R1 Distill Qwen 32B
DeepSeek R1 Distill Qwen 32B is a 32.8B dense qwen2 model from unsloth. Configured context length is 131,072 tokens. License: apache-2.0. Quant sizes are the GGUF file bytes published on Hugging Face when a file exists.
Parameters32.8B
Context128K
Attentiongqa
Licenseapache-2.0
64 layers · 8 KV heads · head dim 128 · GGUF repo · config
Quant file sizes
| Quant | File size | Source |
|---|---|---|
| F16 | 61.0 GB | Hugging Face file bytes |
| Q8_0 | 32.4 GB | Hugging Face file bytes |
| Q6_K | 25.0 GB | Hugging Face file bytes |
| Q5_K_M | 21.7 GB | Hugging Face file bytes |
| Q4_K_M | 18.5 GB | Hugging Face file bytes |
| Q3_K_M | 14.8 GB | Hugging Face file bytes |
| Q2_K | 23.1 GB | Hugging Face file bytes |
On common machines
| Device | Verdict | Quant | Speed |
|---|---|---|---|
| GeForce RTX 4060 Ti 8GB | Partial offload | Q4_K_M | 3.1–4.2 |
| GeForce RTX 3060 12GB | Partial offload | Q4_K_M | 3.9–5.4 |
| GeForce RTX 4060 Ti 16GB | Partial offload | Q4_K_M | 4.7–6.5 |
| GeForce RTX 4090 | Tight fit | Q4_K_M | 25–34 |
| GeForce RTX 5090 | Tight fit | Q6_K | 34–46 |
| Apple M4 Pro | Tight fit | Q3_K_M | 8.3–11 |
| Apple M5 | Won't fit | Q3_K_M | — |
| CPU only · DDR5-5600 | Runs on CPU | Q5_K_M | 1.9–2.6 |
Smallest devices that fit it fully
- RTX 4000 Ada — Tight fit, Q3_K_M, 11–15 tok/s
- Radeon RX 7900 XT — Tight fit, Q3_K_M, 24–33 tok/s
- GeForce RTX 4090 — Tight fit, Q4_K_M, 25–34 tok/s
- GeForce RTX 3090 Ti — Tight fit, Q4_K_M, 25–34 tok/s
- GeForce RTX 3090 — Tight fit, Q4_K_M, 23–32 tok/s
- NVIDIA L4 — Tight fit, Q4_K_M, 7.5–10 tok/s
Run it
Commands for the GeForce RTX 4060 Ti 16GB estimate (Q4_K_M).
llama.cpp
llama-server -hf unsloth/DeepSeek-R1-Distill-Qwen-32B-GGUF:Q4_K_M -c 4096 -ngl 46
No matching name in the Ollama library snapshot.
LM Studio
lms get unsloth/DeepSeek-R1-Distill-Qwen-32B-GGUF # Then load quant Q4_K_M