이 모델의 강점은 무엇인가요?

고급 멀티모달 기능 128K 긴 컨텍스트 처리 용량 Apache 2.0 오픈소스 라이선스

이 모델의 약점은 무엇인가요?

40.0B 연산 자원 부담 특정 분야 특화 제한 실시간 업데이트 능력 부족

어떤 용도에 가장 적합한가요?

의료 데이터의 멀티모달 분석 전문 장문 문서 요약 및 분석 오픈소스 의료 AI 개발

모델 목록으로

Google Research오픈소스

MedGemma 1.5 4B

Name: MedGemma 1.5 4B
Author: Google Research

MedGemma 1.5 4B는 Google Research가 개발한 멀티모달 기반 모델입니다. 약 40.0B의 파라미터 규모를 가지며, 128K의 긴 컨텍스트 윈도우를 지원합니다.

파라미터

40.0B

컨텍스트

128K

라이선스

Apache 2.0

출시일

2026-01-13

API 가격

이 모델의 API 가격 정보는 현재 공개되지 않았습니다

강점

・고급 멀티모달 기능
・128K 긴 컨텍스트 처리 용량
・Apache 2.0 오픈소스 라이선스

약점

・40.0B 연산 자원 부담
・특정 분야 특화 제한
・실시간 업데이트 능력 부족

활용 사례

・의료 데이터의 멀티모달 분석
・전문 장문 문서 요약 및 분석
・오픈소스 의료 AI 개발

심층 분석

Parameters

Updated version of MedGemma 1 4B with expanded capabilities

Release Date

January 2026

MedGemma 1.5 update

New Capabilities

CT/MRI, WSI, longitudinal imaging

3D volume interpretation, whole-slide histopathology, longitudinal chest X-rays

Anatomical Localization

Bounding box support

Can localize anatomical features in chest X-rays

EHR Understanding

Yes

Interprets text-based Electronic Health Record data

Lab Report Extraction

Yes

Extracts structured data from unstructured medical lab reports

License

HAI-DEF Terms of Use

Health AI Developer Foundations license

Base Model

Gemma 3

With medically optimized SigLIP image encoder

강점

・Expands MedGemma 1 4B with CT/MRI 3D imaging, whole-slide histopathology, and longitudinal imaging
・Bounding box anatomical localization in chest X-rays
・Structured data extraction from unstructured lab reports
・EHR text interpretation capability
・Improved accuracy on medical text reasoning vs MedGemma 1 4B
・Runs on mobile/edge hardware

약점

・Only available as 4B multimodal instruction-tuned variant (no larger sizes for 1.5)
・HAI-DEF license more restrictive than Apache 2.0
・Limited to text output (no image generation)
・Requires fine-tuning for optimal clinical performance
・New capabilities may have limited benchmark coverage

경쟁사 비교

Model	Arena	SWE	GPQA	Price
MedGemma 1.5 4B	N/A	N/A	~67% (MedQA est)	Free (open weights)
MedGemma 1 4B	N/A	N/A	64.4% (MedQA)	Free (open weights)
MedGemma 1 27B	N/A	N/A	87.7% (MedQA)	Free (open weights)
BioMistral 7B	N/A	N/A	~60% (MedQA)	Free (open weights)

개요

MedGemma 1.5 4B is the updated version of MedGemma 1 4B, expanding medical AI capabilities to include 3D CT/MRI interpretation, whole-slide histopathology, longitudinal chest X-ray comparison, anatomical bounding box localization, and structured lab report extraction. Released January 2026.

벤치마크 및 성능

Improved accuracy on medical text reasoning and modest improvement on standard 2D image interpretation vs MedGemma 1 4B. New capabilities for 3D imaging, histopathology, and EHR understanding are the main additions.

상세 비교

Direct upgrade from MedGemma 1 4B with expanded modality support. Still 4B parameters so runs on mobile. The 3D imaging and histopathology capabilities are unique at this model size. MedGemma 27B remains more capable for pure medical knowledge tasks.

커뮤니티 평가

Appreciated by medical AI developers for the expanded imaging capabilities. The CT/MRI and histopathology support opens new use cases in radiology and pathology. The bounding box localization feature is particularly useful for clinical workflows.

활용 사례

3D CT/MRI volume interpretation, whole-slide histopathology analysis, longitudinal chest X-ray comparison (current vs historical), anatomical feature localization, lab report data extraction, EHR text interpretation. Ideal for radiology and pathology AI development.