FITMYLLM · JULY 28, 2026 · OPEN METHODOLOGY · COMMUNITY BENCHMARKS
LIVE · UPDATED DAILY
Quantization Options Select your GPU for speed estimates Context length: 4K 8K 16K 32K 64K 128K
Quant Bits VRAM @ 16K Quality IQ3_XXS 3.25 16.8 GB
13.8 + 3.0 KV
low IQ3_XS 3.5 17.8 GB
14.8 + 3.0 KV
low Q3_K_S 3.64 18.4 GB
15.4 + 3.0 KV
low IQ3_M 3.76 18.9 GB
15.9 + 3.0 KV
low Q3_K_M 4 19.9 GB
16.9 + 3.0 KV
low Q3_K_L 4.3 21.1 GB
18.1 + 3.0 KV
moderate IQ4_XS 4.46 21.8 GB
18.8 + 3.0 KV
moderate Q4_K_S 4.67 22.6 GB
19.6 + 3.0 KV
moderate Q4_K_M 4.89 23.5 GB
20.5 + 3.0 KV
good Q5_K_S 5.57 26.3 GB
23.3 + 3.0 KV
good Q5_K_M 5.7 26.9 GB
23.9 + 3.0 KV
good Q6_K 6.56 30.4 GB
27.4 + 3.0 KV
excellent Q8_0 8.5 38.3 GB
35.3 + 3.0 KV
lossless FP16 16 69.1 GB
66.1 + 3.0 KV
lossless
Select your GPU above to see speed estimates and compatibility for each quantization.
Deploying for a team or in production? Size GPUs, cost & scaling in Enterprise → ▸ READY TO RUN THIS? RENT BY THE HOUR
RENT A GPU AND RUN DEEPSEEK-R1-DISTILL-QWEN-32B NOW
Spin up an A100 / H100 / 4090 in ~60s. Pay by the second. Cancel anytime.
Community Ratings Chat Coding Reasoning Creative Vision Roleplay Agentic
Loading ratings...
Run this model IQ3_XXS — 13.8 GB VRAM IQ3_XS — 14.8 GB VRAM Q3_K_S — 15.4 GB VRAM IQ3_M — 15.9 GB VRAM Q3_K_M — 16.9 GB VRAM Q3_K_L — 18.1 GB VRAM IQ4_XS — 18.8 GB VRAM Q4_K_S — 19.6 GB VRAM Q4_K_M — 20.5 GB VRAM Q5_K_S — 23.3 GB VRAM Q5_K_M — 23.9 GB VRAM Q6_K — 27.4 GB VRAM Q8_0 — 35.3 GB VRAM FP16 — 66.1 GB VRAM
Ollama llama.cpp vLLM LM Studio KoboldCpp Jan Docker
▸ Easiest way to get started · Beginners
DOCS ↗ curl -fsSL https://ollama.com/install.sh | shCOPY
$ ollama run deepseek-r1COPY
Tag may need adjustment — check ollama.com/library/deepseek-r1 for available tags.
▸ SETUP GUIDE >_
Auto-setup with fitmyllm CLI Detects your GPU, recommends the best model, downloads it, and starts chatting — zero config. Benchmarks your speed and contributes anonymous data to improve predictions.
Auto-detect GPU Live tok/s in chat Speed benchmarks 9 inference engines
GPUs that can run this model At Q4_K_M quantization. Sorted by minimum VRAM.
Find the best GPU for DeepSeek-R1-Distill-Qwen-32B
Build Hardware for DeepSeek-R1-Distill-Qwen-32B We introduce our first-generation reasoning models, DeepSeek-R1-Zero and DeepSeek-R1.
Read full model card ▸ COLOPHON FITMYLLM · INDEPENDENT · DATA-DRIVEN
FITMYLLM · EST. 2025 · © 2026
RECOMMENDATIONS FROM PUBLISHED MATH, CORRECTED BY THE COMMUNITY — 30.
▸ SPEC SHEET
DeepSeek-R1-Distill-Qwen-32B — 32.8B Dense. ▸ SPECIFICATIONS
PARAMETERS 32.8B
ARCHITECTURE Dense Transformer
CONTEXT LENGTH 128K tokens
CAPABILITIES reasoning, chat
RELEASE DATE 2025-01-20
PROVIDER DeepSeek
FAMILY qwen ▸ VRAM REQUIREMENTS
QUANT BPW VRAM QUALITY IQ3_XXS 3.25 13.8 GB 82% IQ3_XS 3.5 14.8 GB 84% Q3_K_S 3.64 15.4 GB 85% IQ3_M 3.76 15.9 GB 86% Q3_K_M 4 16.9 GB 88% Q3_K_L 4.3 18.1 GB 90% IQ4_XS 4.46 18.8 GB 92% Q4_K_S 4.67 19.6 GB 93% Q4_K_M 4.89 20.5 GB 94% Q5_K_S 5.57 23.3 GB 96% Q5_K_M 5.7 23.9 GB 96% Q6_K 6.56 27.4 GB 97% Q8_0 8.5 35.3 GB 100% FP16 16 66.1 GB 100%
§ 01 BENCHMARK SCORES
HumanEval 87.2
MMLU-PRO 51.9
MATH 62.5
IFEval 83.5
BBH 56.5
GPQA 11.7
MUSR 13.5
MBPP 77.0
BigCodeBench 45.0
Arena Elo 1470.0
GPQA Diamond 61.5
LiveCodeBench 27.0
AIME 63.0
MATH-500 94.1
HLE 5.5
AA Intelligence 17.2
AA Math 63.0
aa_ifbench 22.9
aa_scicode 37.6
aa_lcr 9.7
§ 02 RUN COMMAND
Run DeepSeek-R1-Distill-Qwen-32B locally with Ollama — needs 20.5 GB VRAM at Q4_K_M:
$ ollama run deepseek-r1
§ 03 COMPATIBLE GPUs
30 @ Q4_K_M Feedback 