README.md

---
language: en
library_name: mlx
tags:
- quantized
- mlx
base_model: XiaomiMiMo/MiMo-V2.5
base_model_relation: quantized
pipeline_tag: text-generation
---
# NOTICE
<marquee direction="left" width="400">
    <h1 style="font-size: 60px; color: red;">CURRENTLY UPLOADING...&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;</h1>
</marquee>

**See MiMo-V2.5 in action - [demonstration videos](https://youtube.com/xcreate)**

#### Tested on a M3 Ultra 512GB RAM using [Inferencer app v1.11.1](https://inferencer.com)
- Text inference: ~32.8 tokens/s @ 1000 tokens (debug build)
- Omnimodal inference: Not included in this language model (LM) only version
- Memory usage: ~209 GiB

![Screenshot](https://cdn-uploads.huggingface.co/production/uploads/6930102f5a301d8b708c299c/LGqxJLnVGR8aqslkRZ-v8.jpeg)

<p style="margin-bottom:0px;">
<strong>Q5.8-INF uses the data-agnostic INF method tuned to yield maximum general accuracy within a 256 GiB memory budget</strong>

The perplexity results below show the loss relative to the Q9 edition on our coding test.
</p>
<table style="border-collapse: collapse; text-align:center; margin-top:10px; margin-bottom:0px;">
<thead>
<tr><th>Quantization (bpw)</th><th>Perplexity</th><th>Token Accuracy</th><th>Missed Divergence</th><th>Size</th></tr></tr>
</thead>
<tbody>
<tr><td><strong>Q5.5</strong></td><td>1.07812</td><td>98.25%</td><td>11.45%</td><td>212.35 GB</td></tr>
<tr><td><strong>Q5.8-INF</strong></td><td>1.07031</td><td>99.30%</td><td>10.81%</td><td>225.03 GB</td></tr>
<tr><td><strong>Q6.5</strong></td><td>1.07812</td><td>99.15%</td><td>0.05%</td><td>250.94 GB</td></tr>
<tr><td><strong>Q9 (Base)</strong></td><td>1.07031</td><td>100.00%</td><td>100.00%</td><td>347.43 GB</td></tr>
</tbody>
</table>

- Perplexity: Measures the confidence for predicting base tokens (lower is better)
- Token Accuracy: The percentage of correctly generated base tokens
- Missed Divergence: Measures severity of misses; how much the token was missed by

##### Quantized with a modified version of [MLX](https://github.com/ml-explore/mlx)
##### For more details see [demonstration videos](https://youtube.com/xcreate) or visit [XiaomiMiMo/MiMo-V2.5](https://huggingface.co/XiaomiMiMo/MiMo-V2.5).

## Disclaimer

We are not the creator, originator, or owner of any model listed. Each model is created and provided by third parties. Models may not always be accurate or contextually appropriate. You are responsible for verifying the information before making important decisions. We are not liable for any damages, losses, or issues arising from its use, including data loss or inaccuracies in AI-generated content.