FitMyLLM
Xiaomi/Mixture of Experts

XMiMo V2.5 Pro

MiMo-V2.5 is a native omnimodal model with strong agentic capabilities, supporting text, image, video, and audio understanding within a unified architecture.

chatcodingreasoningmultilingualmathvisionagentictool_use
1023.24B
Parameters (42B active)
1024K
Context length
13
Benchmarks
17
Quantizations
0
Architecture
MoE
Released
2026-04-22
Layers
70
KV Heads
8
Head Dim
192
Family
mimo

Quantization Options

Context length:
QuantBitsVRAM @ 16KQuality
IQ2_XXS2.38
305.6 GB
304.9 + 0.7 KV
low
IQ2_M2.93
376.0 GB
375.3 + 0.7 KV
low
Q2_K3.16
405.4 GB
404.7 + 0.7 KV
low
IQ3_XXS3.25
416.9 GB
416.2 + 0.7 KV
low
IQ3_XS3.5
448.9 GB
448.2 + 0.7 KV
low
Q3_K_S3.64
466.8 GB
466.1 + 0.7 KV
low
IQ3_M3.76
482.2 GB
481.4 + 0.7 KV
low
Q3_K_M4
512.9 GB
512.1 + 0.7 KV
low
Q3_K_L4.3
551.2 GB
550.5 + 0.7 KV
moderate
IQ4_XS4.46
571.7 GB
570.9 + 0.7 KV
moderate
Q4_K_S4.67
598.6 GB
597.8 + 0.7 KV
moderate
Q4_K_M4.89
626.7 GB
625.9 + 0.7 KV
good
Q5_K_S5.57
713.7 GB
712.9 + 0.7 KV
good
Q5_K_M5.7
730.3 GB
729.5 + 0.7 KV
good
Q6_K6.56
840.3 GB
839.5 + 0.7 KV
excellent
Q8_08.5
1088.4 GB
1087.7 + 0.7 KV
lossless
FP1616
2047.7 GB
2047.0 + 0.7 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 ~306 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 MiMo V2.5 Pro now

Spin up an A100 / H100 / 4090 in ~60s. Pay by the second. Cancel anytime.

Community Ratings

Loading ratings...

Benchmarks (13)

Run this model

Easiest way to get started·Beginners
Docs ↗
curl -fsSL https://ollama.com/install.sh | sh
$# MiMo V2.5 Pro is not in Ollama's library, and we know no GGUF build of it

Search Hugging Face for a GGUF of it (https://huggingface.co/models?sort=trending&search=MiMo-V2.5-Pro+GGUF), or use the vLLM tab with the original weights.

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.

pip install fitmyllmthen run fitmyllmLearn more
Auto-detect GPULive tok/s in chatSpeed benchmarks9 inference engines

Find the best GPU for MiMo V2.5 Pro

Build Hardware for MiMo V2.5 Pro
Spec sheet

MiMo V2.5 Pro — 1023.24B MoE.

Specifications
Parameters
1023.24B (42B active)
Architecture
Mixture of Experts
Context Length
1024K tokens
Capabilities
chat, coding, reasoning, multilingual, math, vision, agentic, tool_use
Release Date
2026-04-22
Provider
Xiaomi
Family
mimo
VRAM requirements
QuantBPWVRAMQuality
IQ2_XXS2.38304.9 GB65%
IQ2_M2.93375.3 GB75%
Q2_K3.16404.7 GB78%
IQ3_XXS3.25416.2 GB82%
IQ3_XS3.5448.2 GB84%
Q3_K_S3.64466.1 GB85%
IQ3_M3.76481.4 GB86%
Q3_K_M4512.1 GB88%
Q3_K_L4.3550.5 GB90%
IQ4_XS4.46570.9 GB92%
Q4_K_S4.67597.8 GB93%
Q4_K_M4.89625.9 GB94%
Q5_K_S5.57712.9 GB96%
Q5_K_M5.7729.5 GB96%
Q6_K6.56839.5 GB97%
Q8_08.51087.7 GB100%
FP16162047.0 GB100%
Benchmark scores
GPQA Diamond86.6
HLE33.8
AA Intelligence53.8
AA Coding45.5
IFBench79.9
Terminal-Bench Hard43.2
τ²-Bench94.2
SciCode50.2
AA Long Context73.3
SWE-bench78.9
Terminal-Bench 2.165.2
Terminal-Bench 40.0
EQ-Bench Creative Writing v31493