FITMYLLM · AUGUST 17, 2026 · OPEN METHODOLOGY · COMMUNITY BENCHMARKS
LIVE · UPDATED DAILY
397B
Parameters (17B active)
Quantization Options Select your GPU for speed estimates Context length: 4K 8K 16K 32K 64K 128K 256K
Quant Bits VRAM @ 16K Quality IQ2_XXS 2.38 118.9 GB
118.6 + 0.4 KV
low IQ2_M 2.93 146.2 GB
145.9 + 0.4 KV
low Q2_K 3.16 157.7 GB
157.3 + 0.4 KV
low IQ3_XXS 3.25 162.1 GB
161.8 + 0.4 KV
low IQ3_XS 3.5 174.5 GB
174.2 + 0.4 KV
low Q3_K_S 3.64 181.5 GB
181.1 + 0.4 KV
low IQ3_M 3.76 187.4 GB
187.1 + 0.4 KV
low Q3_K_M 4 199.3 GB
199.0 + 0.4 KV
low Q3_K_L 4.3 214.2 GB
213.9 + 0.4 KV
moderate IQ4_XS 4.46 222.2 GB
221.8 + 0.4 KV
moderate Q4_K_S 4.67 232.6 GB
232.2 + 0.4 KV
moderate Q4_K_M 4.89 243.5 GB
243.2 + 0.4 KV
good Q5_K_S 5.57 277.3 GB
276.9 + 0.4 KV
good Q5_K_M 5.7 283.7 GB
283.4 + 0.4 KV
good Q6_K 6.56 326.4 GB
326.0 + 0.4 KV
excellent Q8_0 8.5 422.7 GB
422.3 + 0.4 KV
lossless FP16 16 794.8 GB
794.5 + 0.4 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 ~119 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 QWEN3.5-397B-A17B 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 — 118.6 GB VRAM IQ2_M — 145.9 GB VRAM Q2_K — 157.3 GB VRAM IQ3_XXS — 161.8 GB VRAM IQ3_XS — 174.2 GB VRAM Q3_K_S — 181.1 GB VRAM IQ3_M — 187.1 GB VRAM Q3_K_M — 199.0 GB VRAM Q3_K_L — 213.9 GB VRAM IQ4_XS — 221.8 GB VRAM Q4_K_S — 232.2 GB VRAM Q4_K_M — 243.2 GB VRAM Q5_K_S — 276.9 GB VRAM Q5_K_M — 283.4 GB VRAM Q6_K — 326.0 GB VRAM Q8_0 — 422.3 GB VRAM FP16 — 794.5 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 qwen3.5:397b-a17bCOPY
Tag may need adjustment — check ollama.com/library/qwen3.5 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 Qwen3.5-397B-A17B
Build Hardware for Qwen3.5-397B-A17B ▸ COLOPHON FITMYLLM · INDEPENDENT · DATA-DRIVEN
FITMYLLM · EST. 2025 · © 2026
RECOMMENDATIONS FROM PUBLISHED MATH, CORRECTED BY THE COMMUNITY — 30.
▸ SPEC SHEET
Qwen3.5-397B-A17B — 397B MoE. ▸ SPECIFICATIONS
PARAMETERS 397B (17B active)
ARCHITECTURE Mixture of Experts
CONTEXT LENGTH 256K tokens
CAPABILITIES chat, coding, reasoning, vision, multilingual
RELEASE DATE 2026-02-01
PROVIDER Alibaba
FAMILY qwen ▸ VRAM REQUIREMENTS
QUANT BPW VRAM QUALITY IQ2_XXS 2.38 118.6 GB 65% IQ2_M 2.93 145.9 GB 75% Q2_K 3.16 157.3 GB 78% IQ3_XXS 3.25 161.8 GB 82% IQ3_XS 3.5 174.2 GB 84% Q3_K_S 3.64 181.1 GB 85% IQ3_M 3.76 187.1 GB 86% Q3_K_M 4 199.0 GB 88% Q3_K_L 4.3 213.9 GB 90% IQ4_XS 4.46 221.8 GB 92% Q4_K_S 4.67 232.2 GB 93% Q4_K_M 4.89 243.2 GB 94% Q5_K_S 5.57 276.9 GB 96% Q5_K_M 5.7 283.4 GB 96% Q6_K 6.56 326.0 GB 97% Q8_0 8.5 422.3 GB 100% FP16 16 794.5 GB 100%
§ 01 BENCHMARK SCORES
MMLU-PRO 87.8
IFEval 92.6
LiveCodeBench 83.6
SWE-bench 76.4
AIME 91.3
GPQA Diamond 88.4
HLE 28.7
AA Intelligence 45.0
AA Coding 41.3
aa_ifbench 78.8
aa_terminal_bench 40.9
aa_tau2 95.6
aa_scicode 42.0
aa_lcr 65.7
§ 02 RUN COMMAND
Run Qwen3.5-397B-A17B locally with Ollama — needs 243.2 GB VRAM at Q4_K_M:
$ ollama run qwen3.5:397b-a17b
§ 03 COMPATIBLE GPUs
5 @ Q4_K_M Feedback 