FITMYLLM · SEPTEMBER 6, 2026 · OPEN METHODOLOGY · COMMUNITY BENCHMARKS
LIVE · UPDATED DAILY
110B
Parameters (12B active)
Quantization Options Select your GPU for speed estimates Context length: 4K 8K 16K 32K 64K 125K
Quant Bits VRAM @ 16K Quality IQ2_XXS 2.38 35.4 GB
33.2 + 2.2 KV
low IQ2_M 2.93 42.9 GB
40.8 + 2.2 KV
low Q2_K 3.16 46.1 GB
43.9 + 2.2 KV
low IQ3_XXS 3.25 47.3 GB
45.2 + 2.2 KV
low IQ3_XS 3.5 50.8 GB
48.6 + 2.2 KV
low Q3_K_S 3.64 52.7 GB
50.5 + 2.2 KV
low IQ3_M 3.76 54.3 GB
52.2 + 2.2 KV
low Q3_K_M 4 57.6 GB
55.5 + 2.2 KV
low Q3_K_L 4.3 61.8 GB
59.6 + 2.2 KV
moderate IQ4_XS 4.46 64.0 GB
61.8 + 2.2 KV
moderate Q4_K_S 4.67 66.9 GB
64.7 + 2.2 KV
moderate Q4_K_M 4.89 69.9 GB
67.7 + 2.2 KV
good Q5_K_S 5.57 79.2 GB
77.1 + 2.2 KV
good Q5_K_M 5.7 81.0 GB
78.9 + 2.2 KV
good Q6_K 6.56 92.8 GB
90.7 + 2.2 KV
excellent Q8_0 8.5 119.5 GB
117.4 + 2.2 KV
lossless FP16 16 222.6 GB
220.5 + 2.2 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 GLM 4.5 AIR 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 — 33.2 GB VRAM IQ2_M — 40.8 GB VRAM Q2_K — 43.9 GB VRAM IQ3_XXS — 45.2 GB VRAM IQ3_XS — 48.6 GB VRAM Q3_K_S — 50.5 GB VRAM IQ3_M — 52.2 GB VRAM Q3_K_M — 55.5 GB VRAM Q3_K_L — 59.6 GB VRAM IQ4_XS — 61.8 GB VRAM Q4_K_S — 64.7 GB VRAM Q4_K_M — 67.7 GB VRAM Q5_K_S — 77.1 GB VRAM Q5_K_M — 78.9 GB VRAM Q6_K — 90.7 GB VRAM Q8_0 — 117.4 GB VRAM FP16 — 220.5 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:110b-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 4.5 Air
Build Hardware for GLM 4.5 Air GLM 4.5 Air — lighter MoE variant balancing speed and capability.
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 4.5 Air — 110B MoE. ▸ SPECIFICATIONS
PARAMETERS 110B (12B active)
ARCHITECTURE Mixture of Experts
CONTEXT LENGTH 125K 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 33.2 GB 65% IQ2_M 2.93 40.8 GB 75% Q2_K 3.16 43.9 GB 78% IQ3_XXS 3.25 45.2 GB 82% IQ3_XS 3.5 48.6 GB 84% Q3_K_S 3.64 50.5 GB 85% IQ3_M 3.76 52.2 GB 86% Q3_K_M 4 55.5 GB 88% Q3_K_L 4.3 59.6 GB 90% IQ4_XS 4.46 61.8 GB 92% Q4_K_S 4.67 64.7 GB 93% Q4_K_M 4.89 67.7 GB 94% Q5_K_S 5.57 77.1 GB 96% Q5_K_M 5.7 78.9 GB 96% Q6_K 6.56 90.7 GB 97% Q8_0 8.5 117.4 GB 100% FP16 16 220.5 GB 100%
§ 01 BENCHMARK SCORES
MMLU-PRO 81.5
GPQA Diamond 73.3
LiveCodeBench 68.4
AIME 80.7
MATH-500 96.5
HLE 6.8
AA Intelligence 23.2
AA Coding 23.8
AA Math 80.7
aa_ifbench 37.6
aa_terminal_bench 20.5
aa_tau2 46.5
aa_scicode 30.6
aa_lcr 43.7
§ 03 COMPATIBLE GPUs
30 @ Q4_K_M Feedback 