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 16.7 GB
13.7 + 3.0 KV
low IQ3_XS 3.5 17.7 GB
14.7 + 3.0 KV
low Q3_K_S 3.64 18.3 GB
15.3 + 3.0 KV
low IQ3_M 3.76 18.8 GB
15.8 + 3.0 KV
low Q3_K_M 4 19.7 GB
16.7 + 3.0 KV
low Q3_K_L 4.3 21.0 GB
18.0 + 3.0 KV
moderate IQ4_XS 4.46 21.6 GB
18.6 + 3.0 KV
moderate Q4_K_S 4.67 22.5 GB
19.5 + 3.0 KV
moderate Q4_K_M 4.89 23.4 GB
20.4 + 3.0 KV
good Q5_K_S 5.57 26.1 GB
23.1 + 3.0 KV
good Q5_K_M 5.7 26.6 GB
23.6 + 3.0 KV
good Q6_K 6.56 30.1 GB
27.1 + 3.0 KV
excellent Q8_0 8.5 38.0 GB
35.0 + 3.0 KV
lossless FP16 16 68.5 GB
65.5 + 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 QWEN 2.5 CODER 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 — 13.7 GB VRAM IQ3_XS — 14.7 GB VRAM Q3_K_S — 15.3 GB VRAM IQ3_M — 15.8 GB VRAM Q3_K_M — 16.7 GB VRAM Q3_K_L — 18.0 GB VRAM IQ4_XS — 18.6 GB VRAM Q4_K_S — 19.5 GB VRAM Q4_K_M — 20.4 GB VRAM Q5_K_S — 23.1 GB VRAM Q5_K_M — 23.6 GB VRAM Q6_K — 27.1 GB VRAM Q8_0 — 35.0 GB VRAM FP16 — 65.5 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 qwen2.5-coder:32b-base-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 2.5 Coder 32B
Build Hardware for Qwen 2.5 Coder 32B Qwen 2.5 Coder 32B — best open coding model. Matches GPT-4 on many code benchmarks.
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 2.5 Coder 32B — 32.5B Dense. ▸ SPECIFICATIONS
PARAMETERS 32.5B
ARCHITECTURE Dense Transformer
CONTEXT LENGTH 128K tokens
CAPABILITIES coding, chat, reasoning
RELEASE DATE 2024-11-12
PROVIDER Alibaba
FAMILY qwen ▸ VRAM REQUIREMENTS
QUANT BPW VRAM QUALITY IQ3_XXS 3.25 13.7 GB 82% IQ3_XS 3.5 14.7 GB 84% Q3_K_S 3.64 15.3 GB 85% IQ3_M 3.76 15.8 GB 86% Q3_K_M 4 16.7 GB 88% Q3_K_L 4.3 18.0 GB 90% IQ4_XS 4.46 18.6 GB 92% Q4_K_S 4.67 19.5 GB 93% Q4_K_M 4.89 20.4 GB 94% Q5_K_S 5.57 23.1 GB 96% Q5_K_M 5.7 23.6 GB 96% Q6_K 6.56 27.1 GB 97% Q8_0 8.5 35.0 GB 100% FP16 16 65.5 GB 100%
§ 01 BENCHMARK SCORES
HumanEval 92.7
MMLU-PRO 50.7
MATH 83.9
IFEval 81.5
BBH 56.4
GPQA 22.7
MUSR 18.5
MBPP 77.0
BigCodeBench 53.2
Arena Elo 1232.0
LiveCodeBench 29.5
AIME 12.0
GPQA Diamond 41.7
HLE 3.8
§ 02 RUN COMMAND
Run Qwen 2.5 Coder 32B locally with Ollama — needs 20.4 GB VRAM at Q4_K_M:
$ ollama run qwen2.5-coder:32b
§ 03 COMPATIBLE GPUs
30 @ Q4_K_M Feedback 