FITMYLLM · AUGUST 17, 2026 · OPEN METHODOLOGY · COMMUNITY BENCHMARKS
LIVE · UPDATED DAILY
300B
Parameters (47B active)
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 92.3 GB
89.7 + 2.5 KV
low IQ2_M 2.93 112.9 GB
110.4 + 2.5 KV
low Q2_K 3.16 121.5 GB
119.0 + 2.5 KV
low IQ3_XXS 3.25 124.9 GB
122.4 + 2.5 KV
low IQ3_XS 3.5 134.3 GB
131.7 + 2.5 KV
low Q3_K_S 3.64 139.5 GB
137.0 + 2.5 KV
low IQ3_M 3.76 144.0 GB
141.5 + 2.5 KV
low Q3_K_M 4 153.0 GB
150.5 + 2.5 KV
low Q3_K_L 4.3 164.3 GB
161.7 + 2.5 KV
moderate IQ4_XS 4.46 170.3 GB
167.7 + 2.5 KV
moderate Q4_K_S 4.67 178.1 GB
175.6 + 2.5 KV
moderate Q4_K_M 4.89 186.4 GB
183.9 + 2.5 KV
good Q5_K_S 5.57 211.9 GB
209.4 + 2.5 KV
good Q5_K_M 5.7 216.8 GB
214.2 + 2.5 KV
good Q6_K 6.56 249.0 GB
246.5 + 2.5 KV
excellent Q8_0 8.5 321.8 GB
319.2 + 2.5 KV
lossless FP16 16 603.0 GB
600.5 + 2.5 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 ~92 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 ERNIE 4.5 300B-A47B 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 — 89.7 GB VRAM IQ2_M — 110.4 GB VRAM Q2_K — 119.0 GB VRAM IQ3_XXS — 122.4 GB VRAM IQ3_XS — 131.7 GB VRAM Q3_K_S — 137.0 GB VRAM IQ3_M — 141.5 GB VRAM Q3_K_M — 150.5 GB VRAM Q3_K_L — 161.7 GB VRAM IQ4_XS — 167.7 GB VRAM Q4_K_S — 175.6 GB VRAM Q4_K_M — 183.9 GB VRAM Q5_K_S — 209.4 GB VRAM Q5_K_M — 214.2 GB VRAM Q6_K — 246.5 GB VRAM Q8_0 — 319.2 GB VRAM FP16 — 600.5 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 ernie:300b-q4_K_MCOPY
Tag may need adjustment — check ollama.com/library/ernie 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 ERNIE 4.5 300B-A47B
Build Hardware for ERNIE 4.5 300B-A47B ▸ COLOPHON FITMYLLM · INDEPENDENT · DATA-DRIVEN
FITMYLLM · EST. 2025 · © 2026
RECOMMENDATIONS FROM PUBLISHED MATH, CORRECTED BY THE COMMUNITY — 30.
▸ SPEC SHEET
ERNIE 4.5 300B-A47B — 300B MoE. ▸ SPECIFICATIONS
PARAMETERS 300B (47B active)
ARCHITECTURE Mixture of Experts
CONTEXT LENGTH 128K tokens
CAPABILITIES chat, coding, reasoning, math
RELEASE DATE 2025-06-28
FAMILY ernie ▸ VRAM REQUIREMENTS
QUANT BPW VRAM QUALITY IQ2_XXS 2.38 89.7 GB 65% IQ2_M 2.93 110.4 GB 75% Q2_K 3.16 119.0 GB 78% IQ3_XXS 3.25 122.4 GB 82% IQ3_XS 3.5 131.7 GB 84% Q3_K_S 3.64 137.0 GB 85% IQ3_M 3.76 141.5 GB 86% Q3_K_M 4 150.5 GB 88% Q3_K_L 4.3 161.7 GB 90% IQ4_XS 4.46 167.7 GB 92% Q4_K_S 4.67 175.6 GB 93% Q4_K_M 4.89 183.9 GB 94% Q5_K_S 5.57 209.4 GB 96% Q5_K_M 5.7 214.2 GB 96% Q6_K 6.56 246.5 GB 97% Q8_0 8.5 319.2 GB 100% FP16 16 600.5 GB 100%
§ 01 BENCHMARK SCORES
HumanEval 92.1
MMLU-PRO 78.4
IFEval 88.0
BBH 94.3
MUSR 69.9
LiveCodeBench 38.8
AIME 54.8
MATH-500 96.4
GPQA Diamond 81.1
HLE 3.5
AA Intelligence 15.0
AA Coding 14.5
AA Math 41.3
aa_ifbench 39.1
aa_terminal_bench 6.1
aa_scicode 31.5
aa_lcr 2.3
§ 03 COMPATIBLE GPUs
13 @ Q4_K_M Feedback