FITMYLLM · SEPTEMBER 6, 2026 · OPEN METHODOLOGY · COMMUNITY BENCHMARKS
LIVE · UPDATED DAILY
357B
Parameters (12B active)
Quantization Options Select your GPU for speed estimates Context length: 4K 8K 16K 32K 64K 128K 195K
Quant Bits VRAM @ 16K Quality IQ2_XXS 2.38 111.0 GB
106.7 + 4.3 KV
low IQ2_M 2.93 135.6 GB
131.2 + 4.3 KV
low Q2_K 3.16 145.8 GB
141.5 + 4.3 KV
low IQ3_XXS 3.25 149.8 GB
145.5 + 4.3 KV
low IQ3_XS 3.5 161.0 GB
156.7 + 4.3 KV
low Q3_K_S 3.64 167.2 GB
162.9 + 4.3 KV
low IQ3_M 3.76 172.6 GB
168.3 + 4.3 KV
low Q3_K_M 4 183.3 GB
179.0 + 4.3 KV
low Q3_K_L 4.3 196.7 GB
192.4 + 4.3 KV
moderate IQ4_XS 4.46 203.8 GB
199.5 + 4.3 KV
moderate Q4_K_S 4.67 213.2 GB
208.9 + 4.3 KV
moderate Q4_K_M 4.89 223.0 GB
218.7 + 4.3 KV
good Q5_K_S 5.57 253.4 GB
249.0 + 4.3 KV
good Q5_K_M 5.7 259.2 GB
254.9 + 4.3 KV
good Q6_K 6.56 297.5 GB
293.2 + 4.3 KV
excellent Q8_0 8.5 384.1 GB
379.8 + 4.3 KV
lossless FP16 16 718.8 GB
714.5 + 4.3 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 ~111 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 GLM 4.6 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 — 106.7 GB VRAM IQ2_M — 131.2 GB VRAM Q2_K — 141.5 GB VRAM IQ3_XXS — 145.5 GB VRAM IQ3_XS — 156.7 GB VRAM Q3_K_S — 162.9 GB VRAM IQ3_M — 168.3 GB VRAM Q3_K_M — 179.0 GB VRAM Q3_K_L — 192.4 GB VRAM IQ4_XS — 199.5 GB VRAM Q4_K_S — 208.9 GB VRAM Q4_K_M — 218.7 GB VRAM Q5_K_S — 249.0 GB VRAM Q5_K_M — 254.9 GB VRAM Q6_K — 293.2 GB VRAM Q8_0 — 379.8 GB VRAM FP16 — 714.5 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 glm:357b-q4_K_MCOPY
Tag may need adjustment — check ollama.com/library/glm 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 GLM 4.6
Build Hardware for GLM 4.6 GLM 4.6 — improved reasoning and tool use over GLM 4.5.
Read full model card ▸ COLOPHON FITMYLLM · INDEPENDENT · DATA-DRIVEN
FITMYLLM · EST. 2025 · © 2026
RECOMMENDATIONS FROM PUBLISHED MATH, CORRECTED BY THE COMMUNITY — 30.
▸ SPEC SHEET
GLM 4.6 — 357B MoE. ▸ SPECIFICATIONS
PARAMETERS 357B (12B active)
ARCHITECTURE Mixture of Experts
CONTEXT LENGTH 195K tokens
CAPABILITIES chat, coding, reasoning, tool_use
RELEASE DATE 2025-08-08
PROVIDER Zhipu AI
FAMILY glm ▸ VRAM REQUIREMENTS
QUANT BPW VRAM QUALITY IQ2_XXS 2.38 106.7 GB 65% IQ2_M 2.93 131.2 GB 75% Q2_K 3.16 141.5 GB 78% IQ3_XXS 3.25 145.5 GB 82% IQ3_XS 3.5 156.7 GB 84% Q3_K_S 3.64 162.9 GB 85% IQ3_M 3.76 168.3 GB 86% Q3_K_M 4 179.0 GB 88% Q3_K_L 4.3 192.4 GB 90% IQ4_XS 4.46 199.5 GB 92% Q4_K_S 4.67 208.9 GB 93% Q4_K_M 4.89 218.7 GB 94% Q5_K_S 5.57 249.0 GB 96% Q5_K_M 5.7 254.9 GB 96% Q6_K 6.56 293.2 GB 97% Q8_0 8.5 379.8 GB 100% FP16 16 714.5 GB 100%
§ 01 BENCHMARK SCORES
MMLU-PRO 78.4
GPQA Diamond 71.9
LiveCodeBench 16.0
AIME 85.3
MATH-500 44.3
HLE 8.9
AA Intelligence 23.4
AA Coding 19.7
AA Math 85.3
aa_ifbench 30.1
aa_terminal_bench 14.4
aa_tau2 31.6
aa_scicode 30.4
aa_lcr 40.3
§ 03 COMPATIBLE GPUs
5 @ Q4_K_M Feedback 