MiMo V2 Flash
Active Parameters
15B
Context Length
256K
Modality
Text
Architecture
Mixture of Experts
KV Cache Rate
192 KB/tok
License
MIT
Release Date
10 Dec 2025
Knowledge Cutoff
Dec 2024
API Pricing (per 1M)
Input: $0.10 · Output: $0.30
Hardware Sizing
Precision
Context
Workload Memory: 607.50 GB
FP16 · 8K context
Base Weights: 603.50 GB
KV Cache: 1.50 GB
Runtime & Graphs: 2.50 GB
Minimum Required Hardware
8x NVIDIA H200 141GB
607.5 / 1015.2 GB usable · TP 8
Determine requirements for datacenter hardware:
Upgrade to Pro
Architecture Diagram
Evaluation Benchmarks
Rank
#110
| Benchmark | Score | Rank |
|---|---|---|
Professional Knowledge | 0.849 | 15 |
Graduate-Level QA | 0.837 | 50 |
Coding Index | 0.50 | 87 |
Web Development | auto 1292 Standard 1330 | 108 103 |
General Text | 1392 | 117 |
Intelligence Index | auto 0.21 Standard 0.16 | 145 193 |
Rankings
Overall Rank
#110
Coding Rank
#103
About MiMo V2 Flash
MiMo V2 Flash is a 309B Mixture-of-Experts model by Xiaomi activating 15B parameters for high-speed reasoning and software engineering. Powered by hybrid sliding-window attention and Multi-Token Prediction, it supports contexts up to 256K tokens.
Technical Specifications
Attention
Attention Structure
Multi-Head Attention
Attention Heads
64
Key-Value Heads
8
Attention Head Dimension
128
Position Embedding
Absolute Position Embedding
RoPE Theta
640,000
Normalization
RMS Normalization
Activation Function
SwigLU
Dimensions
Hidden Dimension Size
4,096
Number of Layers
48
FFN Intermediate Size (Dense)
11,008
Multi-Token Prediction Heads
1
Tokenizer
Vocabulary Size
151,680
Mixture of Experts
Total Expert Parameters
309.0B
Number of Experts
256
Active Experts
8
About MiMo V2
MiMo-V2-Flash is a Mixture-of-Experts (MoE) model with hybrid attention architecture designed for high-speed reasoning and agentic workflows. It features Multi-Token Prediction (MTP) to achieve state-of-the-art performance while significantly reducing inference costs. The model is optimized for long-context modeling and efficient inference.