Model Intelligence Sheet
magnitudedev/LFM2.5-8B-A1B-DSpark-GGUF overview
LFM2.5 8B A1B DSpark GGUF Organization owned copy of the existing GGUF from LiquidAI/LFM2.5 8B A1B DSpark GGUF https://huggingface.co/LiquidAI/LFM2.5 8B A1B DS…
Runs locally from ~340.0 MB disk (4 GB VRAM class GPUs with llama.cpp / guIDE).
Repository Files & Downloads
1 GGUF files detected
Direct downloads for local inference
| File | Type | Quantization | Size | Link |
|---|---|---|---|---|
| LFM2.5-8B-A1B-DSpark-Q8_0.gguf | GGUF | Q8_0 | 340.0 MB | Download |
Model Details
| Model ID | magnitudedev/LFM2.5-8B-A1B-DSpark-GGUF |
|---|---|
| Author | magnitudedev |
| Pipeline | — |
| License | other |
| Base model | LiquidAI/LFM2.5-8B-A1B-DSpark-GGUF |
| Last modified | 2026-10-06T22:04:16.000Z |
Model README
---
license: other
license_name: lfm1.0
license_link: LICENSE
base_model: LiquidAI/LFM2.5-8B-A1B-DSpark-GGUF
tags:
- gguf
- speculative-decoding
- magnitude
inference: false
---
LFM2.5-8B-A1B-DSpark-GGUF
Organization-owned copy of the existing GGUF from LiquidAI/LFM2.5-8B-A1B-DSpark-GGUF, preserving its tensor payload and quantization. This is a speculative drafter and requires its corresponding target model.
Artifact and provenance
- File:
LFM2.5-8B-A1B-DSpark-Q8_0.gguf - Source GGUF revision:
7ba04ee5ff05a4baf2681fe5ddda6d736581ccdf - SHA-256:
c15011522a6956cd30336d968dfad80e47f747495111ca090d766a26b4ecf3f1
Draft metadata
dflash.attention.causal = false: the draft block uses bidirectional attention.dflash.sample_from_anchor = true: proposal selection starts at the anchor row.- Derived from the checkpoint configuration and SGLang’s DSpark configuration and DFlash implementation; verified against reference outputs.
Run magnitudedev/LFM2.5-8B-A1B-DSpark-GGUF with guIDE
Download guIDE — the AI-native code editor with local LLM inference and 69 built-in tools.
Source: Hugging Face · Compare models