FITMYLLM · AUGUST 17, 2026 · OPEN METHODOLOGY · COMMUNITY BENCHMARKS
LIVE · UPDATED DAILY
120B
Parameters (12B active)
Quantization Options Select your GPU for speed estimates Context length: 4K 8K 16K 32K 64K 128K 256K 512K 1M
Quant Bits VRAM @ 16K Quality IQ2_XXS 2.38 39.9 GB
36.2 + 3.8 KV
low IQ2_M 2.93 48.2 GB
44.4 + 3.8 KV
low Q2_K 3.16 51.6 GB
47.9 + 3.8 KV
low IQ3_XXS 3.25 53.0 GB
49.2 + 3.8 KV
low IQ3_XS 3.5 56.7 GB
53.0 + 3.8 KV
low Q3_K_S 3.64 58.8 GB
55.1 + 3.8 KV
low IQ3_M 3.76 60.6 GB
56.9 + 3.8 KV
low Q3_K_M 4 64.2 GB
60.5 + 3.8 KV
low Q3_K_L 4.3 68.7 GB
65.0 + 3.8 KV
moderate IQ4_XS 4.46 71.1 GB
67.4 + 3.8 KV
moderate Q4_K_S 4.67 74.3 GB
70.5 + 3.8 KV
moderate Q4_K_M 4.89 77.6 GB
73.8 + 3.8 KV
good Q5_K_S 5.57 87.8 GB
84.0 + 3.8 KV
good Q5_K_M 5.7 89.7 GB
86.0 + 3.8 KV
good Q6_K 6.56 102.6 GB
98.9 + 3.8 KV
excellent Q8_0 8.5 131.7 GB
128.0 + 3.8 KV
lossless FP16 16 244.2 GB
240.5 + 3.8 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 NEMOTRON 3 SUPER 120B 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 — 36.2 GB VRAM IQ2_M — 44.4 GB VRAM Q2_K — 47.9 GB VRAM IQ3_XXS — 49.2 GB VRAM IQ3_XS — 53.0 GB VRAM Q3_K_S — 55.1 GB VRAM IQ3_M — 56.9 GB VRAM Q3_K_M — 60.5 GB VRAM Q3_K_L — 65.0 GB VRAM IQ4_XS — 67.4 GB VRAM Q4_K_S — 70.5 GB VRAM Q4_K_M — 73.8 GB VRAM Q5_K_S — 84.0 GB VRAM Q5_K_M — 86.0 GB VRAM Q6_K — 98.9 GB VRAM Q8_0 — 128.0 GB VRAM FP16 — 240.5 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 nemotron-3-super:120b-a12b-q4_K_MCOPY
Downloads and runs automatically. Add --verbose for speed stats.
▸ 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 Nemotron 3 Super 120B
Build Hardware for Nemotron 3 Super 120B ▸ COLOPHON FITMYLLM · INDEPENDENT · DATA-DRIVEN
FITMYLLM · EST. 2025 · © 2026
RECOMMENDATIONS FROM PUBLISHED MATH, CORRECTED BY THE COMMUNITY — 30.
▸ SPEC SHEET
Nemotron 3 Super 120B — 120B MoE. ▸ SPECIFICATIONS
PARAMETERS 120B (12B active)
ARCHITECTURE Mixture of Experts
CONTEXT LENGTH 1024K tokens
CAPABILITIES chat, coding, reasoning, tool_use, agentic
RELEASE DATE 2026-03-11
PROVIDER NVIDIA
FAMILY nemotron ▸ VRAM REQUIREMENTS
QUANT BPW VRAM QUALITY IQ2_XXS 2.38 36.2 GB 65% IQ2_M 2.93 44.4 GB 75% Q2_K 3.16 47.9 GB 78% IQ3_XXS 3.25 49.2 GB 82% IQ3_XS 3.5 53.0 GB 84% Q3_K_S 3.64 55.1 GB 85% IQ3_M 3.76 56.9 GB 86% Q3_K_M 4 60.5 GB 88% Q3_K_L 4.3 65.0 GB 90% IQ4_XS 4.46 67.4 GB 92% Q4_K_S 4.67 70.5 GB 93% Q4_K_M 4.89 73.8 GB 94% Q5_K_S 5.57 84.0 GB 96% Q5_K_M 5.7 86.0 GB 96% Q6_K 6.56 98.9 GB 97% Q8_0 8.5 128.0 GB 100% FP16 16 240.5 GB 100%
§ 01 BENCHMARK SCORES
MMLU-PRO 83.7
LiveCodeBench 81.2
SWE-bench 60.5
AIME 90.2
GPQA Diamond 79.2
HLE 19.2
AA Intelligence 36.0
AA Coding 31.2
aa_ifbench 71.5
aa_terminal_bench 28.8
aa_tau2 67.8
aa_scicode 36.0
aa_lcr 60.0
§ 02 RUN COMMAND
Run Nemotron 3 Super 120B locally with Ollama — needs 73.8 GB VRAM at Q4_K_M:
$ ollama run nemotron-3-super:120b
§ 03 COMPATIBLE GPUs
30 @ Q4_K_M Feedback 