LFM2.5 VL 3B heretic needs about 4.2 GB of memory at Q4_K_M (a 1.7 GB file) with an 8K context window, so it fits on a 6 GB GPU. This repository has 15 quants, from 1.1 GB (Q2_K) to 5.4 GB (F16). Longer context needs more: at 32K tokens the KV cache alone adds about 8.1 GB.
LFM2.5 VL 3B heretic – multimodal LLM with vision and text
LFM2.5 VL 3B heretic is a decensored variant of LiquidAI/LFM2.5-VL-3B. It supports both text and image input, with a SigLIP2 NaFlex vision encoder and a LFM2.5-2.6B language backbone. The model can process up to 32,768 context tokens.
Good at
- Vision grounding
- Full‑page OCR
- Tool calling
Limits
- Requires two GGUF files for vision
- No dedicated vision‑only mode
- License type is other
Best for
- Multimodal chat with image queries
- OCR tasks with layout annotation
- Tool‑enabled image analysis
Who it’s for. Developers building on‑device multimodal applications that need both text and image understanding. The model is suitable for edge deployments where a single small model can handle vision and language tasks.
Quants & memory
| Quant | File | Est. memory | vs VRAM | Result | |
|---|---|---|---|---|---|
| Q2_K | 1.1 GB | 3.6 GB | Fits | Get file → | |
| IQ3_S | 1.3 GB | 3.8 GB | Fits | Get file → | |
| Q3_K_S | 1.3 GB | 3.8 GB | Fits | Get file → | |
| Q3_K_M | 1.4 GB | 3.9 GB | Fits | Get file → | |
| Q3_K_L | 1.5 GB | 4.0 GB | Fits | Get file → | |
| IQ4_XS | 1.5 GB | 4.0 GB | Fits | Get file → | |
| Q4_0 | 1.6 GB | 4.1 GB | Fits | Get file → | |
| Q4_K_S | 1.6 GB | 4.1 GB | Fits | Get file → | |
| Q4_K_M | 1.7 GB | 4.2 GB | Fits | Get file → | |
| Q4_1 | 1.7 GB | 4.3 GB | Fits | Get file → | |
| Q5_K_S | 1.9 GB | 4.4 GB | Fits | Get file → | |
| Q5_K_M | 1.9 GB | 4.5 GB | Fits | Get file → | |
| Q6_K | 2.2 GB | 4.7 GB | Fits | Get file → | |
| Q8_0 | 2.9 GB | 5.4 GB | Fits | Get file → | |
| F16 | 5.4 GB | 7.9 GB | Tight | Get file → |
For 8 GB VRAM, Q4_K_M (1.67 GB) or Q5_K_M (1.94 GB) are suitable, estimating 4.2 GB or 4.5 GB of memory at 8 k context. For 12 GB VRAM, Q6_K (2.22 GB) or Q8_0 (2.87 GB) provide higher quality with 4.7 GB or 5.4 GB memory usage. Larger models like F16 (5.4 GB) need 7.9 GB memory and are best for high‑quality inference.
What GPU do you need for LFM2.5 VL 3B heretic?
These figures add the file size, an fp16 KV cache at an 8K context window (worked out from this model’s own header) and about 0.5 GB for the runtime. They are estimates, so leave a little headroom, and expect to use less context or a smaller quant if you are right at the limit.
- 4 GB VRAM: the best quant that fits is IQ4_XS (1.5 GB file, about 4.0 GB in use).
- 6 GB VRAM: the best quant that fits is Q8_0 (2.9 GB file, about 5.4 GB in use).
- 8 GB VRAM: the best quant that fits is F16 (5.4 GB file, about 7.9 GB in use).
Which LFM2.5 VL 3B heretic quant should you choose?
Q4_K_M is the usual starting point for LFM2.5 VL 3B heretic: it balances file size and quality well. Here it is a 1.7 GB file.
The Q8_0 file is 2.9 GB, about 1.7 times the size of Q4_K_M. It keeps the weights closer to the original, which only matters if you have the memory to spare.
If memory is tight, the smallest file is Q2_K at 1.1 GB, with the quality trade-offs that come with fewer bits. Use it only when memory is very tight. Expect a clear drop in quality, which is usually worse on small models and easier to live with on large ones.
New to quant names? Read GGUF quantization explained.
Specifications
Architecture from the GGUF header
Source
Frequently asked questions about LFM2.5 VL 3B heretic
How much VRAM does LFM2.5 VL 3B heretic need?
About 4.2 GB at Q4_K_M with an 8K context window. That is the 1.7 GB file, plus the KV cache and roughly 0.5 GB of runtime overhead. Smaller quants need less and longer context needs more.
Can I run LFM2.5 VL 3B heretic on an 8 GB GPU?
Yes, with F16: the 5.4 GB file needs about 7.9 GB at 8K context. Larger quants in this repository do not fit in 8 GB.
What is the file size of LFM2.5 VL 3B heretic GGUF?
It depends on the quant: Q2_K is 1.1 GB, IQ3_S is 1.3 GB, Q3_K_S is 1.3 GB, Q3_K_M is 1.4 GB, Q3_K_L is 1.5 GB, IQ4_XS is 1.5 GB and 9 more.
What context length does LFM2.5 VL 3B heretic support?
The GGUF header lists a trained context length of 32,768 tokens (32K). The KV cache grows with the context you use: about 2.0 GB at 8K and 8.1 GB at 32K.
Where can I download LFM2.5 VL 3B heretic GGUF?
The files are on Hugging Face in the saidutta69/LFM2.5-VL-3B-heretic repository. The download buttons on this page open the file on Hugging Face. ggufmodels.com does not host any model files.
What license does LFM2.5 VL 3B heretic use?
Hugging Face lists the license as other. Licenses can differ between the quantized files and the original model, so check the repository and the base model before commercial use.
Can I run LFM2.5 VL 3B heretic on 8 GB VRAM?
Yes, using Q4_K_M or Q5_K_M quantizations will fit within 8 GB VRAM, with estimated memory usage of 4.2 GB or 4.5 GB at 8 k context.
Do I need both GGUF files for vision?
Yes, the vision tower and projector ship separately as LFM2.5-VL-3B-heretic-mmproj-F16.gguf; both must be loaded for image input.
What languages does the model support?
English, Arabic, Chinese, French, German, Italian, Japanese, Korean, Portuguese, Spanish, Vietnamese, Thai, Indonesian, Hindi, Russian, Polish.