AesSedai/Motif-3-GGUF overview
Requires this PR https://github.com/ggml org/llama.cpp/pull/26298 to run. This repo contains specialized MoE quants for Motif Technologies/Motif 3. The idea be…
Runs locally from ~9.1 MB disk (4 GB VRAM class GPUs with llama.cpp / guIDE).
Repository Files & Downloads
| File | Type | Quantization | Size | Link |
|---|---|---|---|---|
| IQ2_S/Motif-3-IQ2_S-00001-of-00004.gguf | GGUF | IQ2_S | 9.1 MB | Download |
| IQ2_S/Motif-3-IQ2_S-00002-of-00004.gguf | GGUF | IQ2_S | 46.20 GB | Download |
| IQ2_S/Motif-3-IQ2_S-00003-of-00004.gguf | GGUF | IQ2_S | 46.28 GB | Download |
| IQ2_S/Motif-3-IQ2_S-00004-of-00004.gguf | GGUF | IQ2_S | 8.91 GB | Download |
| IQ3_M/Motif-3-IQ3_M-00001-of-00004.gguf | GGUF | IQ3_M | 9.1 MB | Download |
| IQ3_M/Motif-3-IQ3_M-00002-of-00004.gguf | GGUF | IQ3_M | 46.56 GB | Download |
| IQ3_M/Motif-3-IQ3_M-00003-of-00004.gguf | GGUF | IQ3_M | 46.28 GB | Download |
| IQ3_M/Motif-3-IQ3_M-00004-of-00004.gguf | GGUF | IQ3_M | 30.95 GB | Download |
| IQ3_S/Motif-3-IQ3_S-00001-of-00004.gguf | GGUF | IQ3_S | 9.1 MB | Download |
| IQ3_S/Motif-3-IQ3_S-00002-of-00004.gguf | GGUF | IQ3_S | 46.54 GB | Download |
| IQ3_S/Motif-3-IQ3_S-00003-of-00004.gguf | GGUF | IQ3_S | 46.02 GB | Download |
| IQ3_S/Motif-3-IQ3_S-00004-of-00004.gguf | GGUF | IQ3_S | 19.28 GB | Download |
| IQ4_XS/Motif-3-IQ4_XS-00001-of-00005.gguf | GGUF | IQ4_XS | 9.1 MB | Download |
| IQ4_XS/Motif-3-IQ4_XS-00002-of-00005.gguf | GGUF | IQ4_XS | 46.54 GB | Download |
| IQ4_XS/Motif-3-IQ4_XS-00003-of-00005.gguf | GGUF | IQ4_XS | 46.09 GB | Download |
| IQ4_XS/Motif-3-IQ4_XS-00004-of-00005.gguf | GGUF | IQ4_XS | 46.11 GB | Download |
| IQ4_XS/Motif-3-IQ4_XS-00005-of-00005.gguf | GGUF | IQ4_XS | 4.39 GB | Download |
| Q4_K_M/Motif-3-Q4_K_M-00001-of-00005.gguf | GGUF | Q4_K_M | 9.1 MB | Download |
| Q4_K_M/Motif-3-Q4_K_M-00002-of-00005.gguf | GGUF | Q4_K_M | 46.39 GB | Download |
| Q4_K_M/Motif-3-Q4_K_M-00003-of-00005.gguf | GGUF | Q4_K_M | 46.17 GB | Download |
| Q4_K_M/Motif-3-Q4_K_M-00004-of-00005.gguf | GGUF | Q4_K_M | 46.17 GB | Download |
| Q4_K_M/Motif-3-Q4_K_M-00005-of-00005.gguf | GGUF | Q4_K_M | 44.75 GB | Download |
| Q5_K_M/Motif-3-Q5_K_M-00001-of-00006.gguf | GGUF | Q5_K_M | 9.1 MB | Download |
| Q5_K_M/Motif-3-Q5_K_M-00002-of-00006.gguf | GGUF | Q5_K_M | 45.17 GB | Download |
| Q5_K_M/Motif-3-Q5_K_M-00003-of-00006.gguf | GGUF | Q5_K_M | 45.53 GB | Download |
| Q5_K_M/Motif-3-Q5_K_M-00004-of-00006.gguf | GGUF | Q5_K_M | 45.66 GB | Download |
| Q5_K_M/Motif-3-Q5_K_M-00005-of-00006.gguf | GGUF | Q5_K_M | 45.42 GB | Download |
| Q5_K_M/Motif-3-Q5_K_M-00006-of-00006.gguf | GGUF | Q5_K_M | 38.29 GB | Download |
Model Details
Model README
---
base_model:
- Motif-Technologies/Motif-3
---
Requires this PR to run.
This repo contains specialized MoE-quants for Motif-Technologies/Motif-3. The idea being that given the huge size of the FFN tensors compared to the rest of the tensors in the model, it should be possible to achieve a better quality while keeping the overall size of the entire model smaller compared to a similar naive quantization. To that end, the quantization type default is kept in high quality and the FFN UP + FFN GATE tensors are quanted down along with the FFN DOWN tensors.
| Quant | Size | Mixture | PPL | 1-(Mean PPL(Q)/PPL(base)) | KLD |
| :----- | :-------------------- | :------------------------------- | :------------------- | :------------------------ | :------------------ |
| Q8_0 | 314.95 GiB (8.60 BPW) | Q8_0 | 35.367303 ± 0.451587 | +2.8870% | 0.010134 ± 0.000061 |
| Q5_K_M | 220.07 GiB (6.01 BPW) | Q8_0 / Q5_K / Q5_K / Q6_K | 35.336996 ± 0.450859 | +2.7989% | 0.015087 ± 0.000088 |
| Q4_K_M | 183.47 GiB (5.01 BPW) | Q8_0 / Q4_K / Q4_K / Q5_K | 35.635889 ± 0.455428 | +3.6684% | 0.026133 ± 0.000135 |
| IQ4_XS | 143.12 GiB (3.91 BPW) | Q8_0 / IQ3_S / IQ3_S / IQ4_XS | 36.272353 ± 0.463193 | +5.5199% | 0.058303 ± 0.000287 |
| IQ3_M | 123.79 GiB (3.38 BPW) | Q6_K / IQ3_XXS / IQ3_XXS / IQ3_S | 38.364498 ± 0.492873 | +11.6062% | 0.103530 ± 0.000507 |
| IQ3_S | 111.84 GiB (3.05 BPW) | Q6_K / IQ2_S / IQ2_S / IQ3_S | 39.318385 ± 0.502525 | +14.3811% | 0.157063 ± 0.000752 |
| IQ2_S | 101.38 GiB (2.77 BPW) | Q6_K / IQ2_XS / IQ2_XS / IQ3_XXS | 41.701314 ± 0.536307 | +21.3133% | 0.222056 ± 0.001044 |
Run AesSedai/Motif-3-GGUF with guIDE
Download guIDE — the AI-native code editor with local LLM inference and 69 built-in tools.
Source: Hugging Face · Compare models