MedScholar-1.5B-f32-GGUF

MedScholar-1.5B is a compact, instruction-aligned medical question-answering model based on Qwen2.5-1.5B-Instruct, fine-tuned using 1 million samples from the MIRIAD-4.4M dataset, and designed for research and educational exploration of clinical knowledge, not for diagnosis or medical decision-making. The model is trained using the Unsloth framework and QLoRA, employs a minimal QA prompt format, and operates under the Apache-2.0 and ODC-By 1.0 licenses; it is strictly intended for non-clinical, academic use only, with output not to be used for real patient care or advice.

Model Files

File name	Size	Quant Type
MedScholar-1.5B.F32.gguf	6.18 GB	F32
MedScholar-1.5B.BF16.gguf	3.09 GB	BF16
MedScholar-1.5B.F16.gguf	3.09 GB	F16
MedScholar-1.5B.Q8_0.gguf	1.65 GB	Q8_0
MedScholar-1.5B.Q6_K.gguf	1.27 GB	Q6_K
MedScholar-1.5B.Q5_K_M.gguf	1.13 GB	Q5_K_M
MedScholar-1.5B.Q5_K_S.gguf	1.1 GB	Q5_K_S
MedScholar-1.5B.Q4_K_M.gguf	986 MB	Q4_K_M
MedScholar-1.5B.Q4_K_S.gguf	940 MB	Q4_K_S
MedScholar-1.5B.Q3_K_L.gguf	880 MB	Q3_K_L
MedScholar-1.5B.Q3_K_M.gguf	824 MB	Q3_K_M
MedScholar-1.5B.Q3_K_S.gguf	761 MB	Q3_K_S
MedScholar-1.5B.Q2_K.gguf	676 MB	Q2_K

Quants Usage

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

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