FITMYLLM · JULY 28, 2026 · OPEN METHODOLOGY · COMMUNITY BENCHMARKS
LIVE · UPDATED DAILY
228.7B
Parameters (21B active)
Quantization Options Select your GPU for speed estimates Context length: 4K 8K 16K 32K 64K 128K 192K
Quant Bits VRAM @ 16K Quality IQ2_XXS 2.38 71.4 GB
68.5 + 2.9 KV
low IQ2_M 2.93 87.2 GB
84.3 + 2.9 KV
low Q2_K 3.16 93.7 GB
90.8 + 2.9 KV
low IQ3_XXS 3.25 96.3 GB
93.4 + 2.9 KV
low IQ3_XS 3.5 103.5 GB
100.5 + 2.9 KV
low Q3_K_S 3.64 107.5 GB
104.5 + 2.9 KV
low IQ3_M 3.76 110.9 GB
108.0 + 2.9 KV
low Q3_K_M 4 117.7 GB
114.8 + 2.9 KV
low Q3_K_L 4.3 126.3 GB
123.4 + 2.9 KV
moderate IQ4_XS 4.46 130.9 GB
128.0 + 2.9 KV
moderate Q4_K_S 4.67 136.9 GB
134.0 + 2.9 KV
moderate Q4_K_M 4.89 143.2 GB
140.3 + 2.9 KV
good Q5_K_S 5.57 162.6 GB
159.7 + 2.9 KV
good Q5_K_M 5.7 166.3 GB
163.4 + 2.9 KV
good Q6_K 6.56 190.9 GB
188.0 + 2.9 KV
excellent Q8_0 8.5 246.4 GB
243.5 + 2.9 KV
lossless FP16 16 460.8 GB
457.9 + 2.9 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 ~71 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-M2.5 228.7B 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 — 68.5 GB VRAM IQ2_M — 84.3 GB VRAM Q2_K — 90.8 GB VRAM IQ3_XXS — 93.4 GB VRAM IQ3_XS — 100.5 GB VRAM Q3_K_S — 104.5 GB VRAM IQ3_M — 108.0 GB VRAM Q3_K_M — 114.8 GB VRAM Q3_K_L — 123.4 GB VRAM IQ4_XS — 128.0 GB VRAM Q4_K_S — 134.0 GB VRAM Q4_K_M — 140.3 GB VRAM Q5_K_S — 159.7 GB VRAM Q5_K_M — 163.4 GB VRAM Q6_K — 188.0 GB VRAM Q8_0 — 243.5 GB VRAM FP16 — 457.9 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:229b-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-M2.5 228.7B
Build Hardware for MiniMax-M2.5 228.7B ▸ COLOPHON FITMYLLM · INDEPENDENT · DATA-DRIVEN
FITMYLLM · EST. 2025 · © 2026
RECOMMENDATIONS FROM PUBLISHED MATH, CORRECTED BY THE COMMUNITY — 30.
▸ SPEC SHEET
MiniMax-M2.5 228.7B — 228.7B MoE. ▸ SPECIFICATIONS
PARAMETERS 228.7B (21B active)
ARCHITECTURE Mixture of Experts
CONTEXT LENGTH 192K tokens
CAPABILITIES chat
RELEASE DATE 2026-03-10
PROVIDER MiniMax
FAMILY minimax ▸ VRAM REQUIREMENTS
QUANT BPW VRAM QUALITY IQ2_XXS 2.38 68.5 GB 65% IQ2_M 2.93 84.3 GB 75% Q2_K 3.16 90.8 GB 78% IQ3_XXS 3.25 93.4 GB 82% IQ3_XS 3.5 100.5 GB 84% Q3_K_S 3.64 104.5 GB 85% IQ3_M 3.76 108.0 GB 86% Q3_K_M 4 114.8 GB 88% Q3_K_L 4.3 123.4 GB 90% IQ4_XS 4.46 128.0 GB 92% Q4_K_S 4.67 134.0 GB 93% Q4_K_M 4.89 140.3 GB 94% Q5_K_S 5.57 159.7 GB 96% Q5_K_M 5.7 163.4 GB 96% Q6_K 6.56 188.0 GB 97% Q8_0 8.5 243.5 GB 100% FP16 16 457.9 GB 100%
§ 01 BENCHMARK SCORES
MATH 86.3
GPQA 85.2
Arena Elo 1495.0
GPQA Diamond 84.8
HLE 19.1
AA Intelligence 41.9
AA Coding 37.4
aa_ifbench 72.3
aa_terminal_bench 25.8
aa_tau2 86.8
aa_scicode 36.1
aa_lcr 61.0
LiveCodeBench 82.6
AIME 78.3
AA Math 78.3
§ 02 RUN COMMAND
Run MiniMax-M2.5 228.7B locally with Ollama — needs 140.3 GB VRAM at Q4_K_M:
$ ollama run minimax:228b
§ 03 COMPATIBLE GPUs
16 @ Q4_K_M Feedback 