FITMYLLM · SEPTEMBER 6, 2026 · OPEN METHODOLOGY · COMMUNITY BENCHMARKS
LIVE · UPDATED DAILY
Quantization Options Select your GPU for speed estimates Context length: 4K 8K 16K 32K
Quant Bits VRAM @ 16K Quality IQ2_M 2.93 56.9 GB
26.9 + 30.0 KV
low Q2_K 3.16 58.9 GB
28.9 + 30.0 KV
low IQ3_XXS 3.25 59.7 GB
29.7 + 30.0 KV
low IQ3_XS 3.5 62.0 GB
32.0 + 30.0 KV
low Q3_K_S 3.64 63.2 GB
33.2 + 30.0 KV
low IQ3_M 3.76 64.3 GB
34.3 + 30.0 KV
low Q3_K_M 4 66.5 GB
36.5 + 30.0 KV
low Q3_K_L 4.3 69.2 GB
39.2 + 30.0 KV
moderate IQ4_XS 4.46 70.6 GB
40.6 + 30.0 KV
moderate Q4_K_S 4.67 72.5 GB
42.5 + 30.0 KV
moderate Q4_K_M 4.89 74.5 GB
44.5 + 30.0 KV
good Q5_K_S 5.57 80.6 GB
50.6 + 30.0 KV
good Q5_K_M 5.7 81.8 GB
51.8 + 30.0 KV
good Q6_K 6.56 89.5 GB
59.5 + 30.0 KV
excellent Q8_0 8.5 107.0 GB
77.0 + 30.0 KV
lossless FP16 16 174.5 GB
144.5 + 30.0 KV
lossless
Select your GPU above to see speed estimates and compatibility for each quantization.
Too big for a single GPU — plan a multi-GPU deployment
Even the lightest quant needs ~57 GB. Size GPUs, replicas, TCO and scaling for a production setup. Open in Enterprise →
▸ READY TO RUN THIS? RENT BY THE HOUR
RENT A GPU AND RUN QWEN 1.5 72B 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 IQ2_M — 26.9 GB VRAM Q2_K — 28.9 GB VRAM IQ3_XXS — 29.7 GB VRAM IQ3_XS — 32.0 GB VRAM Q3_K_S — 33.2 GB VRAM IQ3_M — 34.3 GB VRAM Q3_K_M — 36.5 GB VRAM Q3_K_L — 39.2 GB VRAM IQ4_XS — 40.6 GB VRAM Q4_K_S — 42.5 GB VRAM Q4_K_M — 44.5 GB VRAM Q5_K_S — 50.6 GB VRAM Q5_K_M — 51.8 GB VRAM Q6_K — 59.5 GB VRAM Q8_0 — 77.0 GB VRAM FP16 — 144.5 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 qwen:72b-chat-q4_K_MCOPY
Downloads and runs automatically. Add --verbose for speed stats.
▸ 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 Qwen 1.5 72B
Build Hardware for Qwen 1.5 72B Qwen 1.5 72B — Alibaba's large model for demanding enterprise tasks.
Read full model card ▸ COLOPHON FITMYLLM · INDEPENDENT · DATA-DRIVEN
FITMYLLM · EST. 2025 · © 2026
RECOMMENDATIONS FROM PUBLISHED MATH, CORRECTED BY THE COMMUNITY — 30.
▸ SPEC SHEET
Qwen 1.5 72B — 72B Dense. ▸ SPECIFICATIONS
PARAMETERS 72B
ARCHITECTURE Dense Transformer
CONTEXT LENGTH 32K tokens
CAPABILITIES chat
RELEASE DATE 2024-02-04
PROVIDER Alibaba
FAMILY qwen ▸ VRAM REQUIREMENTS
QUANT BPW VRAM QUALITY IQ2_M 2.93 26.9 GB 75% Q2_K 3.16 28.9 GB 78% IQ3_XXS 3.25 29.7 GB 82% IQ3_XS 3.5 32.0 GB 84% Q3_K_S 3.64 33.2 GB 85% IQ3_M 3.76 34.3 GB 86% Q3_K_M 4 36.5 GB 88% Q3_K_L 4.3 39.2 GB 90% IQ4_XS 4.46 40.6 GB 92% Q4_K_S 4.67 42.5 GB 93% Q4_K_M 4.89 44.5 GB 94% Q5_K_S 5.57 50.6 GB 96% Q5_K_M 5.7 51.8 GB 96% Q6_K 6.56 59.5 GB 97% Q8_0 8.5 77.0 GB 100% FP16 16 144.5 GB 100%
§ 01 BENCHMARK SCORES
HumanEval 59.1
MMLU-PRO 51.4
MATH 59.8
IFEval 86.4
BBH 61.9
GPQA 16.7
MUSR 11.7
MBPP 61.6
BigCodeBench 38.5
§ 02 RUN COMMAND
Run Qwen 1.5 72B locally with Ollama — needs 44.5 GB VRAM at Q4_K_M:
$ ollama run qwen:72b
§ 03 COMPATIBLE GPUs
30 @ Q4_K_M Feedback 