FITMYLLM · AUGUST 17, 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 IQ2_XXS 2.38 126.9 GB
121.0 + 5.9 KV
low IQ2_M 2.93 154.7 GB
148.8 + 5.9 KV
low Q2_K 3.16 166.4 GB
160.5 + 5.9 KV
low IQ3_XXS 3.25 170.9 GB
165.0 + 5.9 KV
low IQ3_XS 3.5 183.6 GB
177.7 + 5.9 KV
low Q3_K_S 3.64 190.7 GB
184.8 + 5.9 KV
low IQ3_M 3.76 196.7 GB
190.8 + 5.9 KV
low Q3_K_M 4 208.9 GB
203.0 + 5.9 KV
low Q3_K_L 4.3 224.1 GB
218.2 + 5.9 KV
moderate IQ4_XS 4.46 232.2 GB
226.3 + 5.9 KV
moderate Q4_K_S 4.67 242.8 GB
236.9 + 5.9 KV
moderate Q4_K_M 4.89 254.0 GB
248.0 + 5.9 KV
good Q5_K_S 5.57 288.4 GB
282.5 + 5.9 KV
good Q5_K_M 5.7 295.0 GB
289.1 + 5.9 KV
good Q6_K 6.56 338.5 GB
332.6 + 5.9 KV
excellent Q8_0 8.5 436.7 GB
430.8 + 5.9 KV
lossless FP16 16 816.4 GB
810.5 + 5.9 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 ~127 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 HERMES 4 405B 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 — 121.0 GB VRAM IQ2_M — 148.8 GB VRAM Q2_K — 160.5 GB VRAM IQ3_XXS — 165.0 GB VRAM IQ3_XS — 177.7 GB VRAM Q3_K_S — 184.8 GB VRAM IQ3_M — 190.8 GB VRAM Q3_K_M — 203.0 GB VRAM Q3_K_L — 218.2 GB VRAM IQ4_XS — 226.3 GB VRAM Q4_K_S — 236.9 GB VRAM Q4_K_M — 248.0 GB VRAM Q5_K_S — 282.5 GB VRAM Q5_K_M — 289.1 GB VRAM Q6_K — 332.6 GB VRAM Q8_0 — 430.8 GB VRAM FP16 — 810.5 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 other:405b-q4_K_MCOPY
Tag may need adjustment — check ollama.com/library/other 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 Hermes 4 405B
Build Hardware for Hermes 4 405B ▸ COLOPHON FITMYLLM · INDEPENDENT · DATA-DRIVEN
FITMYLLM · EST. 2025 · © 2026
RECOMMENDATIONS FROM PUBLISHED MATH, CORRECTED BY THE COMMUNITY — 30.
▸ SPEC SHEET
Hermes 4 405B — 405B Dense. ▸ SPECIFICATIONS
PARAMETERS 405B
ARCHITECTURE Dense Transformer
CONTEXT LENGTH 128K tokens
CAPABILITIES chat, reasoning, tool_use, agentic
RELEASE DATE 2025-08-25
PROVIDER Nous Research
FAMILY other ▸ VRAM REQUIREMENTS
QUANT BPW VRAM QUALITY IQ2_XXS 2.38 121.0 GB 65% IQ2_M 2.93 148.8 GB 75% Q2_K 3.16 160.5 GB 78% IQ3_XXS 3.25 165.0 GB 82% IQ3_XS 3.5 177.7 GB 84% Q3_K_S 3.64 184.8 GB 85% IQ3_M 3.76 190.8 GB 86% Q3_K_M 4 203.0 GB 88% Q3_K_L 4.3 218.2 GB 90% IQ4_XS 4.46 226.3 GB 92% Q4_K_S 4.67 236.9 GB 93% Q4_K_M 4.89 248.0 GB 94% Q5_K_S 5.57 282.5 GB 96% Q5_K_M 5.7 289.1 GB 96% Q6_K 6.56 332.6 GB 97% Q8_0 8.5 430.8 GB 100% FP16 16 810.5 GB 100%
§ 01 BENCHMARK SCORES
MMLU-PRO 72.9
GPQA Diamond 53.6
LiveCodeBench 54.6
AIME 15.3
HLE 4.2
AA Intelligence 18.0
MATH-500 15.3
§ 03 COMPATIBLE GPUs
5 @ Q4_K_M Feedback