Hikari07jp/MiMo-V2.6-Distill-Qwen-9B-Ablitrated: what it needs to run
Hikari07jp/MiMo-V2.6-Distill-Qwen-9B-Ablitrated is a fine-tune of XiaomiMiMo/MiMo-V2.6-Distill-Qwen-9B. Its configuration matches MiMo-V2.6-Distill-Qwen-9B's on every field that decides memory, so MiMo-V2.6-Distill-Qwen-9B's requirements are this repository's.
Answered by
Geometrypublished data. Both configurations are published files, read at the revisions named below and compared field by field. Not a test that the model loads.MiMo-V2.6-Distill-Qwen-9B (XiaomiMiMo/MiMo-V2.6-Distill-Qwen-9B) — every compared field is stated by both and equal.
Files this repository publishes
| Published file | Format | File size | Memory to run it | Smallest memory size with 5% to spare |
|---|---|---|---|---|
| MiMo-V2.6-Distill-Qwen-9B-Ablitrated-Q4_K_M.gguf | Q4_K_M | 5.63 GB | 6.75 GB | 8 GB |
| MiMo-V2.6-Distill-Qwen-9B-Ablitrated-NVFP4.gguf | not in the file name | 6.07 GB | 7.19 GB | 8 GB |
| MiMo-V2.6-Distill-Qwen-9B-Ablitrated-NVFP4-A8.gguf | not in the file name | 7.11 GB | 8.23 GB | 10 GB |
File size is the repository’s own, as Hugging Face lists it. Memory to run it adds MiMo-V2.6-Distill-Qwen-9B’s cache at 8,192 tokens of context (0.32 GB, F16) and llama.cpp’s runtime reserve (0.80 GB, a default that has not been measured for this file). It is a memory calculation, not a test that the file loads.
MiMo-V2.6-Distill-Qwen-9B: weight-file size by format
| Format | Size | Range | Basis |
|---|---|---|---|
| FP16 | 18.8 GB | 13.2 GB – 19.0 GB | size range |
| Q8_0 | 10.0 GB | 7.00 GB – 12.0 GB | size range |
| Q6_K | 7.73 GB | 5.40 GB – 10.2 GB | size range |
| Q5_K_M | 6.72 GB | 4.53 GB – 10.2 GB | size range |
| Q5_0 | 6.57 GB | 4.53 GB – 10.2 GB | size range |
| Q4_K_M | 5.78 GB | 3.71 GB – 10.2 GB | size range |
| Q4_0 | 5.48 GB | 3.71 GB – 10.2 GB | size range |
| Q3_K_M | 4.71 GB | 2.83 GB – 10.2 GB | size range |
| Q2_K | 3.73 GB | 2.16 GB – 10.2 GB | size range |
These are MiMo-V2.6-Distill-Qwen-9B’s sizes. A model with the same geometry quantizes to the same size, to within any difference in parameter count noted above. Which devices hold each, and how fast they run it, is on MiMo-V2.6-Distill-Qwen-9B’s page.
What was read, and where
- Hikari07jp/MiMo-V2.6-Distill-Qwen-9B-Ablitrated at revision 90eac4a225b2 is tagged by its publisher as a fine-tune of XiaomiMiMo/MiMo-V2.6-Distill-Qwen-9B.
- Hikari07jp/MiMo-V2.6-Distill-Qwen-9B-Ablitrated/config.json at 90eac4a225b2 was compared with XiaomiMiMo/MiMo-V2.6-Distill-Qwen-9B/config.json at 2367e865d009, the revision the catalogue pins: layers and their pattern, widths, head counts, experts, latent and recurrent dimensions, vocabulary, tied embeddings and any vision or audio tower.
- Licence, as tagged
- mit
- Parameters in its safetensors index
- 9.4B
- Task, as tagged
- image-text-to-text
- Last changed on Hugging Face
- 2026-09-24
Publishing this model? A badge for its card
[](https://llmbottleneck.com/hf/Hikari07jp/MiMo-V2.6-Distill-Qwen-9B-Ablitrated)
It states the memory to run the model at Q4_K_M with 8,192 tokens of context — built on this repository’s own Q4_K_M file when it publishes one — and links back to this page, where the evidence is.
Resolve another repository
The same answer as JSON: GET /v1/resolve?repo=Hikari07jp/MiMo-V2.6-Distill-Qwen-9B-Ablitrated, with a free key from /keys.