gghfez/MiMo-V2.5-ikllama-GGUF overview
Model This is a text only GGUF quantization of XiaomiMiMo/MiMo V2.5 with unfused atten q,k,v for compatibility with ik llama.cpp https://github.com/ikawrakow/i…
Runs locally from ~5.7 MB disk (4 GB VRAM class GPUs with llama.cpp / guIDE).
Repository Files & Downloads
| File | Type | Quantization | Size | Link |
|---|---|---|---|---|
| MiMo-V2.5-IQ3_S-00001-of-00004.gguf | GGUF | IQ3_S | 5.7 MB | Download |
| MiMo-V2.5-IQ3_S-00002-of-00004.gguf | GGUF | IQ3_S | 46.43 GB | Download |
| MiMo-V2.5-IQ3_S-00003-of-00004.gguf | GGUF | IQ3_S | 46.53 GB | Download |
| MiMo-V2.5-IQ3_S-00004-of-00004.gguf | GGUF | IQ3_S | 12.36 GB | Download |
| MiMo-V2.5-IQ4_XS-00001-of-00004.gguf | GGUF | IQ4_XS | 5.7 MB | Download |
| MiMo-V2.5-IQ4_XS-00002-of-00004.gguf | GGUF | IQ4_XS | 45.85 GB | Download |
| MiMo-V2.5-IQ4_XS-00003-of-00004.gguf | GGUF | IQ4_XS | 46.04 GB | Download |
| MiMo-V2.5-IQ4_XS-00004-of-00004.gguf | GGUF | IQ4_XS | 44.89 GB | Download |
Model Details
Model README
---
base_model:
- XiaomiMiMo/MiMo-V2.5
tags:
- ik_llama.cpp
---
Model
This is a text-only GGUF quantization of XiaomiMiMo/MiMo-V2.5 with unfused atten q,k,v for compatibility with ik_llama.cpp
Re-uploaded from AesSedai/MiMo-V2.5-GGUF prior to this change
Available quants: IQ4_XS and IQ3_S
From AesSedai's model card:
| Quant | Size | Mixture | PPL | 1-(Mean PPL(Q)/PPL(base)) | KLD |
| :--------- | :--------- | :------- | :------- | :------- | :------- |
| Q8_0 | 306.66 GiB (8.50 BPW) | Unknown / TBD | 5.134769 ± 0.030261 | +0.1230% | 0.012010 ± 0.000150 |
| Q5_K_M | 213.39 GiB (5.92 BPW) | Q8_0 / Q5_K / Q5_K / Q6_K | 5.147654 ± 0.030377 | +0.3743% | 0.014752 ± 0.000240 |
| Q4_K_M | 177.68 GiB (4.93 BPW) | Q8_0 / Q4_K / Q4_K / Q5_K | 5.202785 ± 0.030828 | +1.4493% | 0.020631 ± 0.000251 |
| IQ4_XS | 137.75 GiB (3.82 BPW) | Q8_0 / IQ3_S / IQ3_S / IQ4_XS | 5.272594 ± 0.031193 | +2.8105% | 0.041508 ± 0.000343 |
| IQ3_S | 106.31 GiB (2.95 BPW) | Q6_K / IQ2_S / IQ2_S / IQ3_S | 5.545001 ± 0.033188 | +8.1221% | 0.092415 ± 0.000600 |
Run gghfez/MiMo-V2.5-ikllama-GGUF with guIDE
Download guIDE — the AI-native code editor with local LLM inference and 69 built-in tools.
Source: Hugging Face · Compare models