FITMYLLM · JULY 28, 2026 · OPEN METHODOLOGY · COMMUNITY BENCHMARKS
LIVE · UPDATED DAILY
Quantization Options Select your GPU for speed estimates Context length: 4K 8K 16K 32K 64K 125K
Quant Bits VRAM @ 16K Quality IQ3_XXS 3.25 24.2 GB
19.6 + 4.6 KV
low IQ3_XS 3.5 25.6 GB
21.1 + 4.6 KV
low Q3_K_S 3.64 26.5 GB
21.9 + 4.6 KV
low IQ3_M 3.76 27.2 GB
22.6 + 4.6 KV
low Q3_K_M 4 28.6 GB
24.0 + 4.6 KV
low Q3_K_L 4.3 30.3 GB
25.8 + 4.6 KV
moderate IQ4_XS 4.46 31.3 GB
26.7 + 4.6 KV
moderate Q4_K_S 4.67 32.5 GB
27.9 + 4.6 KV
moderate Q4_K_M 4.89 33.8 GB
29.2 + 4.6 KV
good Q5_K_S 5.57 37.8 GB
33.2 + 4.6 KV
good Q5_K_M 5.7 38.6 GB
34.0 + 4.6 KV
good Q6_K 6.56 43.6 GB
39.0 + 4.6 KV
excellent Q8_0 8.5 55.0 GB
50.4 + 4.6 KV
lossless FP16 16 99.1 GB
94.5 + 4.6 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-H 47B 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 IQ3_XXS — 19.6 GB VRAM IQ3_XS — 21.1 GB VRAM Q3_K_S — 21.9 GB VRAM IQ3_M — 22.6 GB VRAM Q3_K_M — 24.0 GB VRAM Q3_K_L — 25.8 GB VRAM IQ4_XS — 26.7 GB VRAM Q4_K_S — 27.9 GB VRAM Q4_K_M — 29.2 GB VRAM Q5_K_S — 33.2 GB VRAM Q5_K_M — 34.0 GB VRAM Q6_K — 39.0 GB VRAM Q8_0 — 50.4 GB VRAM FP16 — 94.5 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:47b-q4_K_MCOPY
Tag may need adjustment — check ollama.com/library/nemotron 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 Nemotron-H 47B
Build Hardware for Nemotron-H 47B Nemotron-H 47B — NVIDIA's hybrid model pushing STEM benchmarks.
Read full model card ▸ COLOPHON FITMYLLM · INDEPENDENT · DATA-DRIVEN
FITMYLLM · EST. 2025 · © 2026
RECOMMENDATIONS FROM PUBLISHED MATH, CORRECTED BY THE COMMUNITY — 30.
▸ SPEC SHEET
Nemotron-H 47B — 47B Dense. ▸ SPECIFICATIONS
PARAMETERS 47B
ARCHITECTURE Dense Transformer
CONTEXT LENGTH 125K tokens
CAPABILITIES chat, coding, reasoning, math
RELEASE DATE 2025-06-06
PROVIDER NVIDIA
FAMILY nemotron ▸ VRAM REQUIREMENTS
QUANT BPW VRAM QUALITY IQ3_XXS 3.25 19.6 GB 82% IQ3_XS 3.5 21.1 GB 84% Q3_K_S 3.64 21.9 GB 85% IQ3_M 3.76 22.6 GB 86% Q3_K_M 4 24.0 GB 88% Q3_K_L 4.3 25.8 GB 90% IQ4_XS 4.46 26.7 GB 92% Q4_K_S 4.67 27.9 GB 93% Q4_K_M 4.89 29.2 GB 94% Q5_K_S 5.57 33.2 GB 96% Q5_K_M 5.7 34.0 GB 96% Q6_K 6.56 39.0 GB 97% Q8_0 8.5 50.4 GB 100% FP16 16 94.5 GB 100%
§ 01 BENCHMARK SCORES
MATH 96.2
IFEval 84.5
GPQA 65.7
MBPP 91.8
Arena Elo 1312.0
§ 02 RUN COMMAND
Run Nemotron-H 47B locally with Ollama — needs 29.2 GB VRAM at Q4_K_M:
$ ollama run nemotron-h:47b
§ 03 COMPATIBLE GPUs
30 @ Q4_K_M Feedback 