---
license: mit
base_model: XiaomiMiMo/MiMo-V2.6-Flash-RL
library_name: transformers
tags:
- moe
- pruned
- reap
- hope
- mimo_v2
- multimodal
---

# MiMo-V2.6-Flash-REAP50

`XiaomiMiMo/MiMo-V2.6-Flash-RL` with **50% of its routed experts removed** — 256
experts per layer down to **128** — so that it fits and serves on a single
NVIDIA Jetson AGX Thor (117 GiB unified memory). 86.1 GiB across 65 shards.

Vision, audio and video input are preserved; `audio_tokenizer/` ships with the checkpoint.

## GGUF Quantizations (llama.cpp)

Official llama.cpp GGUF quantizations (including native **MXFP4_MOE**, optimal hybrid **Q2_K**, multimodal **mmproj**, and speculative **mtp** draft towers) are available at:
👉 **[patrickbdevaney/MiMo-V2.6-Flash-REAP50-GGUF](https://huggingface.co/patrickbdevaney/MiMo-V2.6-Flash-REAP50-GGUF)**


## How the experts were chosen

Not by activation frequency. Expert saliency was accumulated over a calibration corpus and the
prune set was solved with **HOPE**, which minimises the output error a prune set actually causes
including the *interaction* terms between experts — REAP is the same objective with the
off-diagonal zeroed, and that off-diagonal cannot be recovered after the pass.

| setting | value |
|---|---|
| objective | `hope` |
| saliency criterion | `reap_1_1_1` |
| prune ratio | 0.50, uniform across layers |
| per-domain protection | top 8% of every domain held out of the prune set |
| worst domain retained | 0.9935 (audio) |
| mean retained | 0.9964 |
| HOPE objective pᵀFp | 0.01624 |

Selection is scored per domain and ranked by the **worst** one, never the mean: an average is
how a criterion that destroys one capability outscores one that preserves all of them.

### Retained gated output mass, by calibration domain

| domain | retention |
|---|---|
| audio | 0.9935 |
| image | 0.9937 |
| video | 0.9938 |
| science | 0.9973 |
| math | 0.9974 |
| finance | 0.9974 |
| ballast | 0.9976 |
| code | 0.9983 |
| agentic | 0.9983 |

## Routers

Pruning an expert leaves its router column behind. The routers were refitted by output matching
against the unpruned teacher, routers only, every expert frozen: 47 routers refitted; all 47 kept at the teacher weights (the fit did not beat the baseline).

A refit that failed to beat the untouched baseline was discarded in favour of the baseline, so
no router here is worse than simply slicing the teacher's.

## Limitations

- Calibration was English/Chinese text, code, math, science, finance, agentic traces, and
  image/audio/video captions. Domains outside that mix were not measured.
- The `dflash/` speculative-decoding draft head from the source repo is **not** included: it was
  trained against the unpruned expert set and is not valid for this checkpoint.
- Pruned MoE experts do not come back. This is a lossy, irreversible transform of the base model.

## Provenance

Produced by [patrickbdevaney/xiaomi-2.6-flash-REAP](https://github.com/patrickbdevaney/xiaomi-2.6-flash-REAP)
on a single Jetson AGX Thor. MIT, inherited from the base model — attribution to Xiaomi MiMo.