SC117/LFM2.5-VL-3B-Uncensored-GGUF
<div style="font-family: -apple-system, BlinkMacSystemFont, 'Segoe UI', Roboto, sans-serif; margin-bottom: 24px;"> <div style="background: #f5f5f7; border-radius: 20px; padding: 36px 32px; margin-bottom: 20px; text-align: center; position: relative; overflow: hidden;"> <div style="position: absolute; top: -30px; right: -30px; width: 120px; height: 120px; background: #ddd6fe; border-radius: 50%;"></div> <div style="position: absolute; bottom: -20px; left: 40px; width: 80px; height: 80px; background: #c4b5fd; border-radius: 50%;"></div> <div style="position: absolute; top: 50%; left: -15px; width: 60px; height: 60px; background: #ddd6fe; border-radius: 50%;"></div> <div style="display: inline-flex; gap: 8px; margin-bottom: 16px; position: relative; z-index: 1; flex-wrap: wrap; justify-content: center;"> <span style="background: #6366f1; color: white; font-size: 11px; font-weight: 600; padding: 5px 14px; border-radius: 20px;">ABLITERIX</span> <span style="background: #8b5cf6; color: white; font-size: 11px; font-weight: 600; padding: 5px 14px; border-radius: 20px;">TRIAL 65</span> <span style="background: #0ea5e9; color: white; font-size: 11px; font-weight: 600; padding: 5px 14px; border-radius: 20px;">VISION</span> <span style="background: #10b981; color: white; font-size: 11px; font-weight: 600; padding: 5px 14px; border-radius: 20px;">GGUF + IMATRIX</span> <span style="background: #64748b; color: white; font-size: 11px; font-weight: 600; padding: 5px 14px; border-radius: 20px;">LFM Open 1.0</span> </div> <h1 style="margin: 0 0 8px 0; font-size: 28px; font-weight: 700; color: #1e1b4b; letter-spacing: -0.5px; border: none; position: relative; z-index: 1;">LFM2.5-VL-3B-Uncensored-GGUF</h1> <p style="margin: 8px 0 0 0; font-size: 14px; position: relative; z-index: 1;"><span style="color: #6b7280;">English</span> | <a href="https://huggingface.co/SC117/LFM2.5-VL-3B-Uncensored-GGUF/blob/main/README_zh.md" style="color: #6366f1; text-decoration: none;">π δΈζζζ‘£</a></p> <p style="margin: 8px 0 0 0; font-size: 15px; color: #6b7280; position: relative; z-index: 1;">Uncensored vision-language edge model Β· abliterix Trial 65 merged into official LFM2.5-VL-3B Β· imatrix-calibrated GGUFs + mmproj</p> </div> </div>
<div style="font-family: -apple-system, BlinkMacSystemFont, 'Segoe UI', Roboto, sans-serif; display: flex; flex-direction: column; gap: 20px; margin-bottom: 30px;">
<div style="border: 1px solid #cbd5e1; border-radius: 12px; overflow: hidden; background: #ffffff; box-shadow: 0 2px 4px rgba(0,0,0,0.02);"> <div style="background: linear-gradient(135deg, #6366f1 0%, #8b5cf6 100%); padding: 12px 16px; color: white; font-weight: 700; font-size: 14px; display: flex; align-items: center; gap: 8px;"><span>π</span> About this release</div> <div style="padding: 16px; font-size: 13px; color: #334155; line-height: 1.7;"> <p style="margin: 0 0 12px 0;"><a href="https://huggingface.co/LiquidAI/LFM2.5-VL-3B" target="blank" style="color: #4f46e5; text-decoration: none; font-weight: 700;">LFM2.5-VL-3B</a> is Liquid AI's official <b>3.1B vision-language edge model</b>: the LFM2.5-2.6B hybrid language backbone (30 layers: 22 double-gated short-convolution blocks + 8 GQA, 128K context, 128K vocab) paired with a <b>SigLIP2 NaFlex</b> vision encoder (256Γ256, patch 16) and a language-aligned projector.</p> <p style="margin: 0 0 12px 0;">This release merges the abliterix <b>Trial 65</b> LoRA (from our <a href="https://huggingface.co/SC117/LFM2.5-2.6B-Uncensored" target="blank" style="color: #4f46e5; text-decoration: none; font-weight: 700;">LFM2.5-2.6B-Uncensored</a>, rank-1, alpha = r = 1) into the language-model layers of the official VL checkpoint (<code>q/k/v/outproj</code>, <code>feedforward.w2</code>, <code>conv.outproj</code> β 84 tensors across 30 layers). The vision encoder and projector are <b>untouched</b>, so multimodal capability is fully preserved.</p> <p style="margin: 0 0 12px 0;">Build steps:</p> <ol style="margin: 0 0 12px 0; padding-left: 20px;"> <li style="margin-bottom: 6px;"><b>LoRA merge</b> into the VL checkpoint: <code>W += (B @ A) * (alpha / r)</code>, applied to <code>model.languagemodel.layers.*</code>.</li> <li style="margin-bottom: 6px;"><b>BF16 GGUF conversion</b> of the merged model (llama.cpp <code>lfm2</code> architecture) + <b>mmproj export</b> of the vision tower via <code>converthfto_gguf.py --mmproj --outtype bf16</code>.</li> <li style="margin-bottom: 6px;"><b>imatrix calibration</b> β reused the 401-chunk (β1.6M tokens) importance matrix from the 2.6B release (same architecture, weights closely related).</li> <li style="margin-bottom: 6px;"><b>Quantization</b> with <code>llama-quantize --imatrix</code> into five tiers.</li> </ol> <p style="margin: 0;"><b>License: LFM Open License v1.0</b> (same as the base model).</p> </div> </div>
<div style="border: 1px solid #fdba74; border-radius: 12px; overflow: hidden; background: #fff7ed; box-shadow: 0 2px 4px rgba(0,0,0,0.02);"> <div style="background: linear-gradient(135deg, #f97316 0%, #ea580c 100%); padding: 12px 16px; color: white; font-weight: 700; font-size: 14px; display: flex; align-items: center; gap: 8px;"><span>β οΈ</span> Uncensored notice</div> <div style="padding: 16px; font-size: 13px; color: #334155; line-height: 1.7;"> <p style="margin: 0 0 12px 0;">After merging the <b>Trial 65</b> steering, this model shows a <b>much lower refusal rate</b> on both text and vision-grounded prompts, and can differ substantially from official LFM2.5-VL-3B. Evaluate compliance and safety for your use case; control access and audit as needed.</p> <table style="width: 100%; border-collapse: collapse; font-size: 13px;"><tbody> <tr><td style="padding: 6px 10px; box-shadow: 0 1px 0 0 rgba(128,128,128,0.15); font-weight: bold; color: #334155; background: white;">Refusals (2.6B harmful eval)</td><td style="padding: 6px 10px; box-shadow: 0 1px 0 0 rgba(128,128,128,0.15); color: #4b5563; background: white;"><b>6 / 100</b> (baseline ~90 / 100) β same Trial 65 LoRA, measured on the 2.6B release</td></tr> <tr><td style="padding: 6px 10px; box-shadow: 0 1px 0 0 rgba(128,128,128,0.15); font-weight: bold; color: #334155; background: white;">Spot check (this VL release)</td><td style="padding: 6px 10px; box-shadow: 0 1px 0 0 rgba(128,128,128,0.15); color: #4b5563; background: white;">NSFW fiction / privacy intrusion / crime detail prompts β <b>all answered without refusal</b></td></tr> <tr><td style="padding: 6px 10px; box-shadow: 0 1px 0 0 rgba(128,128,128,0.15); font-weight: bold; color: #334155; background: white;">Vision capability</td><td style="padding: 6px 10px; box-shadow: 0 1px 0 0 rgba(128,128,128,0.15); color: #4b5563; background: white;"><b>Fully preserved</b> β shape/color/OCR description verified after merge</td></tr> <tr><td style="padding: 6px 10px; font-weight: bold; color: #334155; background: white;">Selected trial</td><td style="padding: 6px 10px; color: #4b5563; background: white;">abliterix Trial 65 (rank-1 LoRA, alpha = r = 1)</td></tr> <tr><td style="padding: 6px 10px; font-weight: bold; color: #334155; background: white;">Thinking</td><td style="padding: 6px 10px; color: #4b5563; background: white;">Not available β official VL is trained to answer directly (no <code><think></code> mode)</td></tr> </tbody></table> </div> </div>
<div style="border: 1px solid #cbd5e1; border-radius: 12px; overflow: hidden; background: #ffffff; box-shadow: 0 2px 4px rgba(0,0,0,0.02);"> <div style="background: linear-gradient(135deg, #6366f1 0%, #8b5cf6 100%); padding: 12px 16px; color: white; font-weight: 700; font-size: 14px; display: flex; align-items: center; gap: 8px;"><span>π¦</span> Files</div> <div style="padding: 16px;"> <p style="margin: 0 0 12px 0; font-size: 13px; color: #334155;">Main model (language backbone, <code>lfm2</code> architecture, 128K context) β pick one tier and pair it with the mmproj:</p> <table style="width: 100%; border-collapse: collapse; font-size: 13px;"><thead><tr style="background: rgba(99,102,241,0.06);"> <th style="padding: 8px 10px; border-bottom: 2px solid #6366f1; text-align: left; color: #4f46e5;">File</th> <th style="padding: 8px 10px; border-bottom: 2px solid #6366f1; text-align: left; color: #4f46e5;">Size</th> <th style="padding: 8px 10px; border-bottom: 2px solid #6366f1; text-align: left; color: #4f46e5;">BPW</th> <th style="padding: 8px 10px; border-bottom: 2px solid #6366f1; text-align: left; color: #4f46e5;">Best for</th> </tr></thead><tbody> <tr><td style="padding: 8px 10px; box-shadow: 0 1px 0 0 rgba(128,128,128,0.15); background: white;"><code>-IQ3_XS.gguf</code></td><td style="padding: 8px 10px; box-shadow: 0 1px 0 0 rgba(128,128,128,0.15); background: white;">1.22 GB</td><td style="padding: 8px 10px; box-shadow: 0 1px 0 0 rgba(128,128,128,0.15); background: white;">~3.30</td><td style="padding: 8px 10px; box-shadow: 0 1px 0 0 rgba(128,128,128,0.15); background: white;">Maximum compression (perceptible quality loss on small models)</td></tr> <tr><td style="padding: 8px 10px; box-shadow: 0 1px 0 0 rgba(128,128,128,0.15); background: #eef2ff;"><code>-IQ4XS.gguf</code></td><td style="padding: 8px 10px; box-shadow: 0 1px 0 0 rgba(128,128,128,0.15); background: #eef2ff;">1.52 GB</td><td style="padding: 8px 10px; box-shadow: 0 1px 0 0 rgba(128,128,128,0.15); background: #eef2ff;">~4.25</td><td style="padding: 8px 10px; box-shadow: 0 1px 0 0 rgba(128,128,128,0.15); background: #eef2ff;"><b>Sweet spot β smallest tier with Q4KM-class quality</b></td></tr> <tr><td style="padding: 8px 10px; box-shadow: 0 1px 0 0 rgba(128,128,128,0.15); background: white;"><code>*-Q4KM.gguf</code></td><td style="padding: 8px 10px; box-shadow: 0 1px 0 0 rgba(128,128,128,0.15); background: white;">1.67 GB</td><td style="padding: 8px 10px; box-shadow: 0 1px 0 0 rgba(128,128,128,0.15); background: white;">~4.94</td><td style="padding: 8px 10px; box-shadow: 0 1px 0 0 rgba(128,128,128,0.15); background: white;">Verified everyday default</td></tr> <tr><td style="padding: 8px 10px; box-shadow: 0 1px 0 0 rgba(128,128,128,0.15); background: white;"><code>*-Q6K.gguf</code></td><td style="padding: 8px 10px; box-shadow: 0 1px 0 0 rgba(128,128,128,0.15); background: white;">2.22 GB</td><td style="padding: 8px 10px; box-shadow: 0 1px 0 0 rgba(128,128,128,0.15); background: white;">~6.56</td><td style="padding: 8px 10px; box-shadow: 0 1px 0 0 rgba(128,128,128,0.15); background: white;">Quality-first local use</td></tr> <tr><td style="padding: 8px 10px; background: white;"><code>-Q8_0.gguf</code></td><td style="padding: 8px 10px; background: white;">2.87 GB</td><td style="padding: 8px 10px; background: white;">~8.50</td><td style="padding: 8px 10px; background: white;">Near-lossless (imatrix optional here)</td></tr> <tr><td style="padding: 8px 10px; background: white;"><code>-BF16.gguf</code></td><td style="padding: 8px 10px; background: white;">5.40 GB</td><td style="padding: 8px 10px; background: white;">16.00</td><td style="padding: 8px 10px; background: white;">Lossless baseline (source of all tiers)</td></tr> </tbody></table> <p style="margin: 12px 0 0 0; font-size: 13px; color: #334155;">Vision tower (required, pick one):</p> <table style="width: 100%; border-collapse: collapse; font-size: 13px;"><thead><tr style="background: rgba(16,185,129,0.08);"> <th style="padding: 8px 10px; border-bottom: 2px solid #10b981; text-align: left; color: #047857;">File</th> <th style="padding: 8px 10px; border-bottom: 2px solid #10b981; text-align: left; color: #047857;">Size</th> <th style="padding: 8px 10px; border-bottom: 2px solid #10b981; text-align: left; color: #047857;">Notes</th> </tr></thead><tbody> <tr><td style="padding: 8px 10px; box-shadow: 0 1px 0 0 rgba(128,128,128,0.15); background: white;"><code>mmproj--BF16.gguf</code></td><td style="padding: 8px 10px; box-shadow: 0 1px 0 0 rgba(128,128,128,0.15); background: white;">0.86 GB</td><td style="padding: 8px 10px; box-shadow: 0 1px 0 0 rgba(128,128,128,0.15); background: white;">Full-precision vision tower, recommended default</td></tr> <tr><td style="padding: 8px 10px; box-shadow: 0 1px 0 0 rgba(128,128,128,0.15); background: #ecfdf5;"><code>mmproj--Q80.gguf</code></td><td style="padding: 8px 10px; box-shadow: 0 1px 0 0 rgba(128,128,128,0.15); background: #ecfdf5;">0.58 GB</td><td style="padding: 8px 10px; box-shadow: 0 1px 0 0 rgba(128,128,128,0.15); background: #ecfdf5;">8-bit vision tower for low-memory devices (negligible quality difference, same as official Q80 mmproj)</td></tr> </tbody></table> <p style="margin: 12px 0 0 0; font-size: 12px; color: #64748b;">All main-model tiers are <code>lfm2</code> architecture, 128K context, single-file GGUFs, imatrix-calibrated.</p> </div> </div>
<div style="border: 1px solid #cbd5e1; border-radius: 12px; overflow: hidden; background: #ffffff; box-shadow: 0 2px 4px rgba(0,0,0,0.02);"> <div style="background: linear-gradient(135deg, #6366f1 0%, #8b5cf6 100%); padding: 12px 16px; color: white; font-weight: 700; font-size: 14px; display: flex; align-items: center; gap: 8px;"><span>π‘</span> Why imatrix?</div> <div style="padding: 16px; font-size: 13px; color: #334155; line-height: 1.7;"> <p style="margin: 0;">The importance matrix (401 chunks / β1.6M tokens of mixed conversation, math, and code data, computed on the 2.6B-Uncensored BF16 GGUF) tells the quantizer which weights are sensitive. Since the VL language backbone shares the exact same architecture and near-identical weights, the matrix transfers cleanly (166/166 quantized tensors matched; only <code>tokenembd</code> falls back to plain q6K). K-quants and especially the <code>IQ</code> tiers use it to keep more bits on attention/embedding paths.</p> </div> </div>
<div style="border: 1px solid #cbd5e1; border-radius: 12px; overflow: hidden; background: #ffffff; box-shadow: 0 2px 4px rgba(0,0,0,0.02);"> <div style="background: linear-gradient(135deg, #6366f1 0%, #8b5cf6 100%); padding: 12px 16px; color: white; font-weight: 700; font-size: 14px; display: flex; align-items: center; gap: 8px;"><span>π</span> Usage (llama.cpp)</div> <div style="padding: 16px; font-size: 13px; color: #334155; line-height: 1.7;"> <p style="margin: 0 0 12px 0;">Pair any main-model tier with the mmproj (recent llama.cpp, e.g. b10299+, required for <code>lfm2</code> VL support):</p> <pre style="background: #f8fafc; border: 1px solid #e2e8f0; border-radius: 8px; padding: 12px; font-size: 12px; overflow-x: auto; line-height: 1.6;"><code>llama-server -m LFM2.5-VL-3B-Uncensored-Q4KM.gguf \ --mmproj mmproj-LFM2.5-VL-3B-BF16.gguf \ --ctx-size 8192 --flash-attn on --host 0.0.0.0 --port 8080</code></pre> <p style="margin: 12px 0 0 0;">CLI with an image:</p> <pre style="background: #f8fafc; border: 1px solid #e2e8f0; border-radius: 8px; padding: 12px; font-size: 12px; overflow-x: auto; line-height: 1.6;"><code>llama-llava-cli -m LFM2.5-VL-3B-Uncensored-Q4KM.gguf \ --mmproj mmproj-LFM2.5-VL-3B-BF16.gguf \ -i image.jpg -p "Describe this image." -ngl 99</code></pre> <p style="margin: 12px 0 0 0;">Transformers / vLLM / SGLang users: use the merged BF16 safetensors in the parent repo (to be published).</p> </div> </div>
<div style="border: 1px solid #cbd5e1; border-radius: 12px; overflow: hidden; background: #ffffff; box-shadow: 0 2px 4px rgba(0,0,0,0.02);"> <div style="background: linear-gradient(135deg, #6366f1 0%, #8b5cf6 100%); padding: 12px 16px; color: white; font-weight: 700; font-size: 14px; display: flex; align-items: center; gap: 8px;"><span>π§ </span> Behavior notes</div> <div style="padding: 16px; font-size: 13px; color: #334155; line-height: 1.7;"> <ul style="margin: 0; padding-left: 20px;"> <li style="margin-bottom: 6px;"><b>No thinking mode.</b> Unlike the 2.6B text model, official LFM2.5-VL-3B is trained to answer directly for low-latency edge tasks; the <code><think></code> tags are not generated. Keep <code>maxnewtokens</code> modest.</li> <li style="margin-bottom: 6px;"><b>Vision strengths</b> (per official model card): near-real-time object detection, OCR with layout annotation, document/chart understanding, on-device translation.</li> <li style="margin-bottom: 6px;"><b>Chat template</b>: ChatML-like with <code><image></code> placeholder, same as official. Recommended sampling: <code>temperature 0.2, topk 50, repetitionpenalty 1.0</code>.</li> </ul> </div> </div>
<div style="border: 1px solid #cbd5e1; border-radius: 12px; overflow: hidden; background: #ffffff; box-shadow: 0 2px 4px rgba(0,0,0,0.02);"> <div style="background: linear-gradient(135deg, #6366f1 0%, #8b5cf6 100%); padding: 12px 16px; color: white; font-weight: 700; font-size: 14px; display: flex; align-items: center; gap: 8px;"><span>π</span> Benchmarks (official model card)</div> <div style="padding: 16px; font-size: 13px; color: #334155; line-height: 1.7;"> <p style="margin: 0 0 12px 0;">Values below are taken from the <b>official LiquidAI model card</b> for LFM2.5-VL-3B (not re-measured on this release):</p> <table style="width: 100%; border-collapse: collapse; font-size: 13px;"><tbody> <tr><td style="padding: 6px 10px; box-shadow: 0 1px 0 0 rgba(128,128,128,0.15); font-weight: bold; color: #334155; background: white;">MME</td><td style="padding: 6px 10px; box-shadow: 0 1px 0 0 rgba(128,128,128,0.15); color: #4b5563; background: white;">73.1</td><td style="padding: 6px 10px; box-shadow: 0 1px 0 0 rgba(128,128,128,0.15); font-weight: bold; color: #334155; background: white;">ChartQA</td><td style="padding: 6px 10px; box-shadow: 0 1px 0 0 rgba(128,128,128,0.15); color: #4b5563; background: white;">81.3</td></tr> <tr><td style="padding: 6px 10px; box-shadow: 0 1px 0 0 rgba(128,128,128,0.15); font-weight: bold; color: #334155; background: white;">MMStar</td><td style="padding: 6px 10px; box-shadow: 0 1px 0 0 rgba(128,128,128,0.15); color: #4b5563; background: white;">63.3</td><td style="padding: 6px 10px; box-shadow: 0 1px 0 0 rgba(128,128,128,0.15); font-weight: bold; color: #334155; background: white;">OCRBenchv2 (EN)</td><td style="padding: 6px 10px; box-shadow: 0 1px 0 0 rgba(128,128,128,0.15); color: #4b5563; background: white;">47.5</td></tr> <tr><td style="padding: 6px 10px; box-shadow: 0 1px 0 0 rgba(128,128,128,0.15); font-weight: bold; color: #334155; background: white;">RealWorldQA</td><td style="padding: 6px 10px; box-shadow: 0 1px 0 0 rgba(128,128,128,0.15); color: #4b5563; background: white;">73.1</td><td style="padding: 6px 10px; box-shadow: 0 1px 0 0 rgba(128,128,128,0.15); font-weight: bold; color: #334155; background: white;">MathVista</td><td style="padding: 6px 10px; box-shadow: 0 1px 0 0 rgba(128,128,128,0.15); color: #4b5563; background: white;">68.5</td></tr> <tr><td style="padding: 6px 10px; font-weight: bold; color: #334155; background: white;">CountBenchQA</td><td style="padding: 6px 10px; color: #4b5563; background: white;">87.3</td><td style="padding: 6px 10px; font-weight: bold; color: #334155; background: white;">POPE</td><td style="padding: 6px 10px; color: #4b5563; background: white;">88.7</td></tr> </tbody></table> </div> </div>
</div>
<div style="font-family: -apple-system, BlinkMacSystemFont, 'Segoe UI', Roboto, sans-serif; font-size: 12px; color: #64748b; text-align: center; margin-top: 20px;"> Community derivative (behavior edit + quantized GGUF release). <b>Not an official Liquid AI release.</b> Use at your own risk; follow local law and the LFM Open License v1.0. </div>