Qwen2-VL-OCR-2B-Instruct-GGUF [ VL / OCR ]

The Qwen2-VL-OCR-2B-Instruct model is a fine-tuned version of Qwen/Qwen2-VL-2B-Instruct, tailored for tasks that involve Optical Character Recognition (OCR), image-to-text conversion, math problem solving with LaTeX formatting and Messy Handwriting OCR. This model integrates a conversational approach with visual and textual understanding to handle multi-modal tasks effectively.

---

Model Files (Qwen2-VL-OCR-2B-Instruct, GGUF)

File Name	Size	Quantization	Format	Description
Qwen2-VL-OCR-2B-Instruct.f16.gguf	3.09 GB	FP16	GGUF	Full precision (float16)
Qwen2-VL-OCR-2B-Instruct.Q2_K.gguf	676 MB	Q2_K	GGUF	2-bit quantized
Qwen2-VL-OCR-2B-Instruct.Q3_K_L.gguf	880 MB	Q3_K_L	GGUF	3-bit quantized (K L variant)
Qwen2-VL-OCR-2B-Instruct.Q3_K_M.gguf	824 MB	Q3_K_M	GGUF	3-bit quantized (K M variant)
Qwen2-VL-OCR-2B-Instruct.Q3_K_S.gguf	761 MB	Q3_K_S	GGUF	3-bit quantized (K S variant)
Qwen2-VL-OCR-2B-Instruct.Q4_K_M.gguf	986 MB	Q4_K_M	GGUF	4-bit quantized (K M variant)
Qwen2-VL-OCR-2B-Instruct.Q4_K_S.gguf	940 MB	Q4_K_S	GGUF	4-bit quantized (K S variant)
Qwen2-VL-OCR-2B-Instruct.Q5_K_M.gguf	1.13 GB	Q5_K_M	GGUF	5-bit quantized (K M variant)
Qwen2-VL-OCR-2B-Instruct.Q5_K_S.gguf	1.1 GB	Q5_K_S	GGUF	5-bit quantized (K S variant)
Qwen2-VL-OCR-2B-Instruct.Q6_K.gguf	1.27 GB	Q6_K	GGUF	6-bit quantized
Qwen2-VL-OCR-2B-Instruct.Q8_0.gguf	1.65 GB	Q8_0	GGUF	8-bit quantized

---

i1 Quantized Variants

File Name	Size	Quantization	Description
Qwen2-VL-OCR-2B-Instruct.i1-IQ1_M.gguf	464 MB	i1-IQ1_M	i1 1-bit medium
Qwen2-VL-OCR-2B-Instruct.i1-IQ1_S.gguf	437 MB	i1-IQ1_S	i1 1-bit small
Qwen2-VL-OCR-2B-Instruct.i1-IQ2_M.gguf	601 MB	i1-IQ2_M	i1 2-bit medium
Qwen2-VL-OCR-2B-Instruct.i1-IQ2_S.gguf	564 MB	i1-IQ2_S	i1 2-bit small
Qwen2-VL-OCR-2B-Instruct.i1-IQ2_XS.gguf	550 MB	i1-IQ2_XS	i1 2-bit extra small
Qwen2-VL-OCR-2B-Instruct.i1-IQ2_XXS.gguf	511 MB	i1-IQ2_XXS	i1 2-bit extra extra small
Qwen2-VL-OCR-2B-Instruct.i1-IQ3_M.gguf	777 MB	i1-IQ3_M	i1 3-bit medium
Qwen2-VL-OCR-2B-Instruct.i1-IQ3_S.gguf	762 MB	i1-IQ3_S	i1 3-bit small
Qwen2-VL-OCR-2B-Instruct.i1-IQ3_XS.gguf	732 MB	i1-IQ3_XS	i1 3-bit extra small
Qwen2-VL-OCR-2B-Instruct.i1-IQ3_XXS.gguf	669 MB	i1-IQ3_XXS	i1 3-bit extra extra small
Qwen2-VL-OCR-2B-Instruct.i1-IQ4_NL.gguf	936 MB	i1-IQ4_NL	i1 4-bit with no-layernorm quantization
Qwen2-VL-OCR-2B-Instruct.i1-IQ4_XS.gguf	896 MB	i1-IQ4_XS	i1 4-bit extra small
Qwen2-VL-OCR-2B-Instruct.i1-Q4_0.gguf	938 MB	i1-Q4_0	i1 4-bit traditional quant
Qwen2-VL-OCR-2B-Instruct.i1-Q4_1.gguf	1.02 GB	i1-Q4_1	i1 4-bit traditional variant

---

Metadata

File Name	Size	Description
.gitattributes	3.37 kB	Git LFS tracking file
config.json	34 B	Config placeholder
README.md	672 B	Model readme

---

Quants Usage

(sorted by size, not necessarily quality. IQ-quants are often preferable over similar sized non-IQ quants)

Link	Type	Size/GB	Notes
GGUF	Q2_K	0.4
GGUF	Q3_K_S	0.5
GGUF	Q3_K_M	0.5	lower quality
GGUF	Q3_K_L	0.5
GGUF	IQ4_XS	0.6
GGUF	Q4_K_S	0.6	fast, recommended
GGUF	Q4_K_M	0.6	fast, recommended
GGUF	Q5_K_S	0.6
GGUF	Q5_K_M	0.7
GGUF	Q6_K	0.7	very good quality
GGUF	Q8_0	0.9	fast, best quality
GGUF	f16	1.6	16 bpw, overkill

Here is a handy graph by ikawrakow comparing some lower-quality quant types (lower is better):