license: other base_model: blockblockblock/LFM2.5-8B-A1B-uncensored-abliterated base_model_relation: quantized quantized_by: blockblockblock library_name: exllamav3 pipeline_tag: text-generation tags:
- exl3
- exllamav3
- quantized
- mixture-of-experts quantization_format: exl3 bits_per_weight: 4.5
LFM2.5 · 8B · A1B · uncensored · abliterated
EXL3 · 4.5 bpw · 5.4 GB · Mixture‑of‑Experts · 24 layers × 32 experts
base model quantized by collection
[!NOTE] An ExLlamaV3 build of
blockblockblock/LFM2.5-8B-A1B-uncensored-abliteratedat 4.5 bits per weight. See Quants for sibling repos at other bit‑widths or browse the collection.
Quants
| BPW | Head bits | Calibration rows | Size | Status |
|---|---|---|---|---|
| 4.0 | 8 | 250 | 4.9 GB | link |
| 4.5 | 8 | 250 | 5.4 GB | this repo |
| 5.0 | 8 | 250 | 6.0 GB | link |
| 6.0 | 8 | 250 | 7.0 GB | link |
| 8.0 | 8 | 250 | 9.0 GB | link |
Inference
Download
pip install -U huggingface_hub
hf download \
blockblockblock/LFM2.5-8B-A1B-uncensored-abliterated-exl3-4.5bpw \
--local-dir ./LFM2.5-8B-A1B-uncensored-abliterated-exl3-4.5bpw
| Setting | Value |
|---|---|
| Format | EXL3 |
| Bits per weight | 4.5 |
| Head bits | 8 |
| Calibration rows | 250 |
| Codebook | MCG |
| Out‑scales | always |
| Parallel mode | enabled (MoE expert batching) |
Loaded automatically by every ExLlamaV3 loader; reproduced here for searchability.
License & use
[!IMPORTANT] Use and license follow the base model. Quantization adds no additional restrictions. Refer to the upstream repository for terms, citation, and safety documentation.