đ SpaceThinker-Qwen2.5VL-3B Quantized Model
This project provides quantized versions of the SpaceThinker-Qwen2.5VL-3B model, offering various quantization types with different sizes and qualities. It enables efficient usage of the model in different scenarios.
đ Quick Start
If you're new to using this quantized model, refer to the following sections for detailed information on its background, usage, and available quantized versions.
⨠Features
- Multiple Quantization Types: Offers a wide range of quantization types, sorted by size, providing options for different resource requirements and quality preferences.
- Visual Comparison: Includes a graph comparing some lower - quality quant types, helping users make informed decisions.
- External References: Provides links to external resources such as Artefact2's thoughts and model request FAQs.
đ Documentation
About
The model is a weighted/imatrix quant of SpaceThinker - Qwen2.5VL - 3B. Static quants are available at mradermacher/SpaceThinker - Qwen2.5VL - 3B - GGUF.
Property |
Details |
Base Model |
remyxai/SpaceThinker-Qwen2.5VL-3B |
Datasets |
remyxai/SpaceThinker |
Language |
en |
Library Name |
transformers |
License |
apache - 2.0 |
Quantized By |
mradermacher |
Tags |
remyx, qwen2.5 - vl, spatial - reasoning, multimodal, vlm, vqasynth, thinking, reasoning, test - time - compute, robotics, embodied - ai, quantitative - spatial - reasoning, distance - estimation, visual - question - answering |
Usage
If you are unsure how to use GGUF files, refer to one of TheBloke's READMEs for more details, including on how to concatenate multi - part files.
Provided Quants
The following table lists the available quantized versions, sorted by size (not necessarily quality. IQ - quants are often preferable over similar sized non - IQ quants):
Link |
Type |
Size/GB |
Notes |
GGUF |
i1 - IQ1_S |
0.9 |
for the desperate |
GGUF |
i1 - IQ1_M |
1.0 |
mostly desperate |
GGUF |
i1 - IQ2_XXS |
1.0 |
|
GGUF |
i1 - IQ2_XS |
1.1 |
|
GGUF |
i1 - IQ2_M |
1.2 |
|
GGUF |
i1 - Q2_K_S |
1.3 |
very low quality |
GGUF |
i1 - Q2_K |
1.4 |
IQ3_XXS probably better |
GGUF |
i1 - IQ3_XXS |
1.4 |
lower quality |
GGUF |
i1 - IQ3_XS |
1.5 |
|
GGUF |
i1 - Q3_K_S |
1.6 |
IQ3_XS probably better |
GGUF |
i1 - IQ3_S |
1.6 |
beats Q3_K* |
GGUF |
i1 - IQ3_M |
1.6 |
|
GGUF |
i1 - Q3_K_M |
1.7 |
IQ3_S probably better |
GGUF |
i1 - Q3_K_L |
1.8 |
IQ3_M probably better |
GGUF |
i1 - IQ4_XS |
1.8 |
|
GGUF |
i1 - IQ4_NL |
1.9 |
prefer IQ4_XS |
GGUF |
i1 - Q4_0 |
1.9 |
fast, low quality |
GGUF |
i1 - Q4_K_S |
1.9 |
optimal size/speed/quality |
GGUF |
i1 - Q4_K_M |
2.0 |
fast, recommended |
GGUF |
i1 - Q4_1 |
2.1 |
|
GGUF |
i1 - Q5_K_S |
2.3 |
|
GGUF |
i1 - Q5_K_M |
2.3 |
|
GGUF |
i1 - Q6_K |
2.6 |
practically like static Q6_K |
Here is a handy graph by ikawrakow comparing some lower - quality quant types (lower is better):

And here are Artefact2's thoughts on the matter:
Artefact2's Gist
FAQ / Model Request
See mradermacher/model_requests for some answers to questions you might have and/or if you want some other model quantized.
Thanks
I thank my company, nethype GmbH, for letting me use its servers and providing upgrades to my workstation to enable this work in my free time. Additional thanks to @nicoboss for giving me access to his private supercomputer, enabling me to provide many more imatrix quants, at much higher quality, than I would otherwise be able to.
đ License
This project is licensed under the Apache - 2.0 license.