LFM2.5-VL-3B is Liquid AI's official 3.1B vision-language edge model: 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 SigLIP2 NaFlex vision encoder (256×256, patch 16) and a language-aligned projector.

This release merges the abliterix Trial 65 LoRA (from our LFM2.5-2.6B-Uncensored, rank-1, alpha = r = 1) into the language-model layers of the official VL checkpoint (q/k/v/out_proj, feed_forward.w2, conv.out_proj — 84 tensors across 30 layers). The vision encoder and projector are untouched, so multimodal capability is fully preserved.

Build steps:

LoRA merge into the VL checkpoint: W += (B @ A) * (alpha / r), applied to model.language_model.layers.*.
BF16 GGUF conversion of the merged model (llama.cpp lfm2 architecture) + mmproj export of the vision tower via convert_hf_to_gguf.py --mmproj --outtype bf16.
imatrix calibration — reused the 401-chunk (≈1.6M tokens) importance matrix from the 2.6B release (same architecture, weights closely related).
Quantization with llama-quantize --imatrix into five tiers.

License: LFM Open License v1.0 (same as the base model).

⚠️ Uncensored notice

After merging the Trial 65 steering, this model shows a much lower refusal rate 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.

Refusals (2.6B harmful eval)	6 / 100 (baseline ~90 / 100) — same Trial 65 LoRA, measured on the 2.6B release
Spot check (this VL release)	NSFW fiction / privacy intrusion / crime detail prompts → all answered without refusal
Vision capability	Fully preserved — shape/color/OCR description verified after merge
Selected trial	abliterix Trial 65 (rank-1 LoRA, alpha = r = 1)
Thinking	Not available — official VL is trained to answer directly (no `<think>` mode)

📦 Files

Main model (language backbone, lfm2 architecture, 128K context) — pick one tier and pair it with the mmproj:

File	Size	BPW	Best for
`*-IQ3_XS.gguf`	1.22 GB	~3.30	Maximum compression (perceptible quality loss on small models)
`*-IQ4_XS.gguf`	1.52 GB	~4.25	Sweet spot — smallest tier with Q4_K_M-class quality
`*-Q4_K_M.gguf`	1.67 GB	~4.94	Verified everyday default
`*-Q6_K.gguf`	2.22 GB	~6.56	Quality-first local use
`*-Q8_0.gguf`	2.87 GB	~8.50	Near-lossless (imatrix optional here)
`*-BF16.gguf`	5.40 GB	16.00	Lossless baseline (source of all tiers)

Vision tower (required, pick one):

File	Size	Notes
`mmproj-*-BF16.gguf`	0.86 GB	Full-precision vision tower, recommended default
`mmproj-*-Q8_0.gguf`	0.58 GB	8-bit vision tower for low-memory devices (negligible quality difference, same as official Q8_0 mmproj)

All main-model tiers are lfm2 architecture, 128K context, single-file GGUFs, imatrix-calibrated.

💡 Why imatrix?

The importance matrix (401 chunks / ≈1.6M 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 token_embd falls back to plain q6_K). K-quants and especially the IQ tiers use it to keep more bits on attention/embedding paths.

🚀 Usage (llama.cpp)

Pair any main-model tier with the mmproj (recent llama.cpp, e.g. b10299+, required for lfm2 VL support):

llama-server -m LFM2.5-VL-3B-Uncensored-Q4_K_M.gguf \
  --mmproj mmproj-LFM2.5-VL-3B-BF16.gguf \
  --ctx-size 8192 --flash-attn on --host 0.0.0.0 --port 8080

CLI with an image:

llama-llava-cli -m LFM2.5-VL-3B-Uncensored-Q4_K_M.gguf \
  --mmproj mmproj-LFM2.5-VL-3B-BF16.gguf \
  -i image.jpg -p "Describe this image." -ngl 99

Transformers / vLLM / SGLang users: use the merged BF16 safetensors in the parent repo (to be published).

🧠 Behavior notes

No thinking mode. Unlike the 2.6B text model, official LFM2.5-VL-3B is trained to answer directly for low-latency edge tasks; the <think> tags are not generated. Keep max_new_tokens modest.
Vision strengths (per official model card): near-real-time object detection, OCR with layout annotation, document/chart understanding, on-device translation.
Chat template: ChatML-like with <image> placeholder, same as official. Recommended sampling: temperature 0.2, top_k 50, repetition_penalty 1.0.

📊 Benchmarks (official model card)

Values below are taken from the official LiquidAI model card for LFM2.5-VL-3B (not re-measured on this release):

MME	73.1	ChartQA	81.3
MMStar	63.3	OCRBenchv2 (EN)	47.5
RealWorldQA	73.1	MathVista	68.5
CountBenchQA	87.3	POPE	88.7

Community derivative (behavior edit + quantized GGUF release). Not an official Liquid AI release. Use at your own risk; follow local law and the LFM Open License v1.0.

SC117/LFM2.5-VL-3B-Uncensored-GGUF

LFM2.5-VL-3B-Uncensored-GGUF

SC117/
LFM2.5-VL-3B-Uncensored-GGUF