MiMo V2.6 Flash, GGUF
MiMo V2.6 Flash is a native omnimodal Mixture-of-Experts model designed for coding agents, computer use, cybersecurity, and long-horizon tasks across text, image, video, and audio. It integrates vision and audio encoders with Multi-Token Prediction for efficient agentic inference.
This repository holds Vultr's GGUF quantizations of it. They are hybrid quantizations: rather than applying one precision to the whole network, each layer's precision is chosen from a measured importance score, so the bits go where they change the output most.
Model Overview
| Attribute | Value |
|---|---|
| Base model | XiaomiMiMo/MiMo-V2.6-Flash-RL |
| Architecture | mimo2, Mixture-of-Experts |
| Parameters | 309B total |
| Active parameters | 15B |
| Layers | 48 |
| Hidden size | 4,096 |
| Attention heads | 64 |
| Key-value heads | 8 |
| Experts | 256 routed, 8 activated per token |
| Context length | 1M tokens |
| Vocabulary | 152,576 |
| Attention layout | 39 Sliding Window Attention layers, 9 Global Attention layers |
| License | MIT |
| Attribute | Value |
|---|---|
| Base model | XiaomiMiMo/MiMo-V2.6-Flash-RL |
| Architecture | mimo2, Mixture-of-Experts |
| Parameters | 309B total |
| Active parameters | 15B |
| Layers | 48 |
| Hidden size | 4,096 |
| Attention heads | 64 |
| Key-value heads | 8 |
| Experts | 256 routed, 8 activated per token |
| Context length | 1M tokens |
| Vocabulary | 152,576 |
| Attention layout | 39 Sliding Window Attention layers, 9 Global Attention layers |
| License | MIT |
Quantizations
| File | Type | Size | KLD vs original |
|---|---|---|---|
| MiMo-V2.6-Flash-RL-MXFP4-*.gguf (2 parts) | MXFP4 | 167.37 GB | 0.0086 |
| MiMo-V2.6-Flash-RL-IQ3_M.gguf | IQ3_M | 148.62 GB | 0.0527 |
| MiMo-V2.6-Flash-RL-IQ3_S.gguf | IQ3_S | 146.76 GB | 0.0544 |
| File | Type | Size | KLD vs original |
|---|---|---|---|
| MiMo-V2.6-Flash-RL-MXFP4-*.gguf (2 parts) | MXFP4 | 167.37 GB | 0.0086 |
| MiMo-V2.6-Flash-RL-IQ3_M.gguf | IQ3_M | 148.62 GB | 0.0527 |
| MiMo-V2.6-Flash-RL-IQ3_S.gguf | IQ3_S | 146.76 GB | 0.0544 |
The vision encoder is in MiMo-V2.6-Flash-RL-mmproj-f16.gguf (2.74 GB). llama.cpp downloads it automatically with the command under Run, so image input works without extra steps.
KL divergence measures how far a quantized model's next-token probabilities drift from those of the original model as released (its routed experts are already in 4-bit MXFP4), averaged over every token of the evaluation text. The lower the value, the closer the quantized model is to the original; zero would mean identical output.
Evaluation
Every variant in this repository was evaluated against the original model as released (its routed experts are already in 4-bit MXFP4) for perplexity, KL divergence, top-token agreement, HellaSwag and WinoGrande, under identical conditions. The full results are maintained on the model's page in the Vultr Model Library:
https://docs.vultr.com/models/xiaomi-mimo-v26/mimo-v2.6-flash?tab=quantized
That page is the reference for these files. The summary above is an excerpt from it.
Run
llama-server -hf vultr/MiMo-V2.6-Flash-RL-GGUF:MXFP4
The model's chat template is embedded in every file, so no extra prompt formatting is needed.
The repository also includes mtp-MiMo-V2.6-Flash-RL-Q8_0.gguf, the model's multi-token prediction (MTP) head, for optional speculative decoding with any quant above:
llama-server -hf vultr/MiMo-V2.6-Flash-RL-GGUF:MXFP4 --spec-type draft-mtp
License
These files are derived from XiaomiMiMo/MiMo-V2.6-Flash-RL and are distributed under the same license, MIT. The model is the work of Xiaomi MiMo; Vultr produced the quantizations.