FITMYLLM · AUGUST 17, 2026 · OPEN METHODOLOGY · COMMUNITY BENCHMARKS
LIVE · UPDATED DAILY
141B
Parameters (39B active)
0
Quantization Options Select your GPU for speed estimates Context length: 4K 8K 16K 32K 64K
Quant Bits VRAM @ 16K Quality IQ2_XXS 2.38 45.1 GB
42.4 + 2.6 KV
low IQ2_M 2.93 54.8 GB
52.1 + 2.6 KV
low Q2_K 3.16 58.8 GB
56.2 + 2.6 KV
low IQ3_XXS 3.25 60.4 GB
57.8 + 2.6 KV
low IQ3_XS 3.5 64.8 GB
62.2 + 2.6 KV
low Q3_K_S 3.64 67.3 GB
64.6 + 2.6 KV
low IQ3_M 3.76 69.4 GB
66.8 + 2.6 KV
low Q3_K_M 4 73.6 GB
71.0 + 2.6 KV
low Q3_K_L 4.3 78.9 GB
76.3 + 2.6 KV
moderate IQ4_XS 4.46 81.7 GB
79.1 + 2.6 KV
moderate Q4_K_S 4.67 85.4 GB
82.8 + 2.6 KV
moderate Q4_K_M 4.89 89.3 GB
86.7 + 2.6 KV
good Q5_K_S 5.57 101.3 GB
98.7 + 2.6 KV
good Q5_K_M 5.7 103.6 GB
101.0 + 2.6 KV
good Q6_K 6.56 118.7 GB
116.1 + 2.6 KV
excellent Q8_0 8.5 152.9 GB
150.3 + 2.6 KV
lossless FP16 16 285.1 GB
282.5 + 2.6 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 WIZARDLM 2 8X22B 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_XXS — 42.4 GB VRAM IQ2_M — 52.1 GB VRAM Q2_K — 56.2 GB VRAM IQ3_XXS — 57.8 GB VRAM IQ3_XS — 62.2 GB VRAM Q3_K_S — 64.6 GB VRAM IQ3_M — 66.8 GB VRAM Q3_K_M — 71.0 GB VRAM Q3_K_L — 76.3 GB VRAM IQ4_XS — 79.1 GB VRAM Q4_K_S — 82.8 GB VRAM Q4_K_M — 86.7 GB VRAM Q5_K_S — 98.7 GB VRAM Q5_K_M — 101.0 GB VRAM Q6_K — 116.1 GB VRAM Q8_0 — 150.3 GB VRAM FP16 — 282.5 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 wizardlm2:8x22bCOPY
Tag may need adjustment — check ollama.com/library/wizardlm2 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 WizardLM 2 8x22B
Build Hardware for WizardLM 2 8x22B WizardLM 2 8x22B — Microsoft's MoE model for complex instructions.
Read full model card ▸ COLOPHON FITMYLLM · INDEPENDENT · DATA-DRIVEN
FITMYLLM · EST. 2025 · © 2026
RECOMMENDATIONS FROM PUBLISHED MATH, CORRECTED BY THE COMMUNITY — 30.
▸ SPEC SHEET
WizardLM 2 8x22B — 141B MoE. ▸ SPECIFICATIONS
PARAMETERS 141B (39B active)
ARCHITECTURE Mixture of Experts
CONTEXT LENGTH 64K tokens
CAPABILITIES chat, reasoning
RELEASE DATE 2024-04-15
PROVIDER WizardLM
FAMILY mistral ▸ VRAM REQUIREMENTS
QUANT BPW VRAM QUALITY IQ2_XXS 2.38 42.4 GB 65% IQ2_M 2.93 52.1 GB 75% Q2_K 3.16 56.2 GB 78% IQ3_XXS 3.25 57.8 GB 82% IQ3_XS 3.5 62.2 GB 84% Q3_K_S 3.64 64.6 GB 85% IQ3_M 3.76 66.8 GB 86% Q3_K_M 4 71.0 GB 88% Q3_K_L 4.3 76.3 GB 90% IQ4_XS 4.46 79.1 GB 92% Q4_K_S 4.67 82.8 GB 93% Q4_K_M 4.89 86.7 GB 94% Q5_K_S 5.57 98.7 GB 96% Q5_K_M 5.7 101.0 GB 96% Q6_K 6.56 116.1 GB 97% Q8_0 8.5 150.3 GB 100% FP16 16 282.5 GB 100%
§ 01 BENCHMARK SCORES
MMLU-PRO 50.7
MATH 49.5
IFEval 84.0
BBH 52.7
GPQA 24.9
MUSR 17.2
GPQA Diamond 17.6
§ 02 RUN COMMAND
Run WizardLM 2 8x22B locally with Ollama — needs 86.7 GB VRAM at Q4_K_M:
$ ollama run wizardlm2:8x22b
§ 03 COMPATIBLE GPUs
30 @ Q4_K_M Feedback 