FITMYLLM · AUGUST 17, 2026 · OPEN METHODOLOGY · COMMUNITY BENCHMARKS
LIVE · UPDATED DAILY
427.04B
Parameters (23B 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 2.38 128.9 GB
127.5 + 1.4 KV
low IQ2_M 2.93 158.3 GB
156.9 + 1.4 KV
low Q2_K 3.16 170.6 GB
169.2 + 1.4 KV
low IQ3_XXS 3.25 175.4 GB
174.0 + 1.4 KV
low IQ3_XS 3.5 188.7 GB
187.3 + 1.4 KV
low Q3_K_S 3.64 196.2 GB
194.8 + 1.4 KV
low IQ3_M 3.76 202.6 GB
201.2 + 1.4 KV
low Q3_K_M 4 215.4 GB
214.0 + 1.4 KV
low Q3_K_L 4.3 231.4 GB
230.0 + 1.4 KV
moderate IQ4_XS 4.46 240.0 GB
238.6 + 1.4 KV
moderate Q4_K_S 4.67 251.2 GB
249.8 + 1.4 KV
moderate Q4_K_M 4.89 262.9 GB
261.5 + 1.4 KV
good Q5_K_S 5.57 299.2 GB
297.8 + 1.4 KV
good Q5_K_M 5.7 306.2 GB
304.8 + 1.4 KV
good Q6_K 6.56 352.1 GB
350.7 + 1.4 KV
excellent Q8_0 8.5 455.6 GB
454.2 + 1.4 KV
lossless FP16 16 856.0 GB
854.6 + 1.4 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 ~129 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 MINIMAX-M3 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 — 127.5 GB VRAM IQ2_M — 156.9 GB VRAM Q2_K — 169.2 GB VRAM IQ3_XXS — 174.0 GB VRAM IQ3_XS — 187.3 GB VRAM Q3_K_S — 194.8 GB VRAM IQ3_M — 201.2 GB VRAM Q3_K_M — 214.0 GB VRAM Q3_K_L — 230.0 GB VRAM IQ4_XS — 238.6 GB VRAM Q4_K_S — 249.8 GB VRAM Q4_K_M — 261.5 GB VRAM Q5_K_S — 297.8 GB VRAM Q5_K_M — 304.8 GB VRAM Q6_K — 350.7 GB VRAM Q8_0 — 454.2 GB VRAM FP16 — 854.6 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 minimax:427b-q4_K_MCOPY
Tag may need adjustment — check ollama.com/library/minimax 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 MiniMax-M3
Build Hardware for MiniMax-M3 ▸ COLOPHON FITMYLLM · INDEPENDENT · DATA-DRIVEN
FITMYLLM · EST. 2025 · © 2026
RECOMMENDATIONS FROM PUBLISHED MATH, CORRECTED BY THE COMMUNITY — 30.
▸ SPEC SHEET
MiniMax-M3 — 427.04B MoE. ▸ SPECIFICATIONS
PARAMETERS 427.04B (23B active)
ARCHITECTURE Mixture of Experts
CONTEXT LENGTH 1024K tokens
CAPABILITIES chat, coding, reasoning, multilingual, vision, math, agentic, tool_use
RELEASE DATE 2026-06-01
PROVIDER MiniMax
FAMILY minimax ▸ VRAM REQUIREMENTS
QUANT BPW VRAM QUALITY IQ2_XXS 2.38 127.5 GB 65% IQ2_M 2.93 156.9 GB 75% Q2_K 3.16 169.2 GB 78% IQ3_XXS 3.25 174.0 GB 82% IQ3_XS 3.5 187.3 GB 84% Q3_K_S 3.64 194.8 GB 85% IQ3_M 3.76 201.2 GB 86% Q3_K_M 4 214.0 GB 88% Q3_K_L 4.3 230.0 GB 90% IQ4_XS 4.46 238.6 GB 92% Q4_K_S 4.67 249.8 GB 93% Q4_K_M 4.89 261.5 GB 94% Q5_K_S 5.57 297.8 GB 96% Q5_K_M 5.7 304.8 GB 96% Q6_K 6.56 350.7 GB 97% Q8_0 8.5 454.2 GB 100% FP16 16 854.6 GB 100%
§ 01 BENCHMARK SCORES
GPQA Diamond 92.9
HLE 39.0
AA Intelligence 45.4
AA Coding 58.6
aa_ifbench 82.9
aa_terminal_bench 42.4
aa_tau2 88.9
aa_scicode 45.4
aa_lcr 80.3
§ 03 COMPATIBLE GPUs
5 @ Q4_K_M Feedback 