DeepSeek-R1-Distill-Qwen-32B
32BDeepSeek
R1 distilled to 32B. Near-frontier reasoning on a single 24GB card (Q3/Q4).
131K
Max Context
3
Quant Variants
GGUF Q4_K_M
Best Quality
+2.6%
Perplexity rise (best build)
Quantization Variants
Per-quant VRAM, quality loss, and inference speed on RTX 4090
Measured = run on this site · Estimated = not run here: calculated, or a figure this site has not verified · Community = public reports
| Format | Level | BPW | VRAM | PPL Loss | Speed | Source | Actions |
|---|---|---|---|---|---|---|---|
| GGUF | Q3_K_M | 3.87 | 17.5 GB | 7.2% | 50 tok/s | Estimated | |
| GGUF | Q4_K_M | 4.85 | 21.7 GB | 2.6% | 42 tok/s | Estimated | |
| EXL2 | 3.5bpw | 3.5 | 16.0 GB | 4.5% | 65 tok/s | Estimated |
Which file each size is for
A file size is not a memory figure: the VRAM column is the calculator's total at 4K context — weights plus KV cache plus a 10% activation buffer. Units differ on purpose: file sizes are decimal GB, as Hugging Face lists them; the VRAM column is GiB (1024³ bytes) with 2% added to the weights, so 7.12 GB of file counts as about 6.8 GiB there. A vision projector (mmproj), where the repo ships one, is a separate download and is in neither number.
- GGUF Q3_K_MSized at the calculator's generic rate for this level; no file was checked.
- GGUF Q4_K_MSized at the calculator's generic rate for this level; no file was checked.
- EXL2 3.5bpwSized at the calculator's generic rate for this level; no file was checked.
Similar models
Compare with DeepSeek-R1-Distill-Qwen-14BDeepSeek-R1-Distill-Qwen-14B
DeepSeek
R1 reasoning distilled into 14B. Huge community interest; excellent chain-of-thought.
DeepSeek-V2-Lite Chat
DeepSeek
MoE general model (~2.4B active). Long context and strong multilingual chat. Its attention is MLA: a GGUF converted by current llama.cpp (tensors attn_k_b / attn_v_b) caches about 1 GB at 32K, which is what the estimates here assume. An older conversion with a single attn_kv_b tensor caches full keys and values instead — about 8.4 GB at 32K.
DeepSeek-R1-Distill-Qwen-7B
DeepSeek
R1 reasoning in a 7B footprint. Best value for 8–12GB VRAM CoT experiments.
DeepSeek-V3
DeepSeek
DeepSeek-V3 frontier MoE (~37B active / 671B total). MLA + FP8; multi-node GPU cluster required at Q4. MLA cache: a GGUF converted by current llama.cpp (attn_k_b / attn_v_b) holds about 2 GB at 32K, which the estimates here assume; an older conversion with a single attn_kv_b tensor caches full keys and values — about 150 GB at 32K.
Running DeepSeek-R1-Distill-Qwen-32B locally
At Q4_K_M and 4K of context, DeepSeek-R1-Distill-Qwen-32B needs about 21.7 GB — 18.7 GB of weights, 1.00 GB of KV cache and a 2.0 GB activation buffer. The smallest cards in this index that clear that comfortably are 32 GB ones — 4 of them, from the RTX 5090 up — and 26 of the 78 cards here do. These are calculated figures, not measurements: the estimate stops counting a card as comfortable at 88% of its VRAM, which is roughly the room a desktop session needs.
What longer context costs
Going from 4K to 32K adds about 7.00 GB, taking the total to 29.4 GB. Weights do not move with context — only the KV cache does, and it grows linearly, so this is the number to watch when planning for long documents. The model's native window is 128K; holding all of it at Q4_K_M would need about 56 GB.
Which build to download
This index tracks 2 formats for it — GGUF, EXL2 — across 3 levels. Q4_K_M carries the lowest published perplexity loss at 2.6%. The fastest level listed is EXL2 3.5bpw at about 65 tok/s on an RTX 4090, batch 1 — an estimate, not a run on this site's hardware. GGUF runs on llama.cpp and Ollama across NVIDIA, AMD and Apple silicon; AWQ and GPTQ target vLLM on CUDA and ROCm; EXL2 is ExLlamaV2 — CUDA only, and now archived, so local use rather than serving.
Common questions
- How much VRAM does DeepSeek-R1-Distill-Qwen-32B need?
- About 21.7 GB at Q4_K_M with 4K of context and batch 1, rising to roughly 29.4 GB at 32K. That figure is weights plus KV cache plus a 10% activation buffer, calculated from the model's architecture rather than measured on a card.
- Will DeepSeek-R1-Distill-Qwen-32B run on a 32GB GPU?
- Yes — at Q4_K_M and 4K context it needs about 21.7 GB, which leaves 10.3 GB spare on a RTX 5090, or on any of the other 32 GB cards here (4 in all). That is the smallest size in this index that clears it comfortably; 26 of 78 do.
- Which quantization of DeepSeek-R1-Distill-Qwen-32B should I use?
- Q4_K_M has the lowest published quality loss (2.6%), and Q4_K_M is the level most people run. All 3 levels in the index are Q3_K_M, Q4_K_M, EXL2 3.5bpw.
Where this model fits
Sized at Q4_K_M with a 4K context window, smallest card first. Comfortable means the estimate uses at most 88% of the memory.
- Runs comfortably on
- RTX 509032GB · +10.3Arc Pro B70 32G32GB · +10.3Radeon AI PRO R9700 32G32GB · +10.3Instinct MI100 32G32GB · +10.3Mac M3 Pro 36G36GB · +5.3
- Tight but possible
- RTX 309024GBRadeon RX 7900 XTX24GBRTX 409024GB
- Compared here
- GGUF vs EXL2
- Try it yourself
- Size it in the calculatorGenerate the run command
How to actually run this
Deployment guides for this model and this class of hardware.
This page's figures change when the model or the runtime does.Last updated 2026-11-07 RSS → /feed.xml