AI NEWS.ccAI는 지금, 미래는 업데이트 중
♡ 저장한 도구 0
AI 모델 라이브러리/OpenGVLab/InternVL3_5-4B-HF
모델 정보

InternVL3_5-4B-HF

OpenGVLab · InternVL3_5-4B-HF

We introduce *InternVL3.5*, a new family of open-source multimodal models that significantly advances versatility, reasoning capability, and inference efficiency along the InternVL series. A key innovation is the *Cascade Reinforcement Learning (Cascade RL)* framework, which enhances reasoning through a two-stage process: offline RL for stable convergence and online RL for refined alignment. This coarse-to-fine training strategy leads to substantial improvements on downstream reasoning tasks, e.g., MMMU and MathVista. To optimize efficiency, we propose a *Visual Resolution Router (ViR)* that dynamically adjusts the resolution of visual tokens without compromising performance. Coupled with ViR, our Decoupled *Vision-Language Deployment (DvD)* strategy separates the vision encoder and language model across different GPUs, effectively balancing computational load. These contributions collectively enable InternVL3.5 to achieve up to a +16.0\% gain in overall reasoning performance and a 4.05 \\(\times\\) inference speedup compared to its predecessor, i.e., InternVL3. In addition, InternVL3.5 supports novel capabilities such as GUI interaction and embodied agency. Notably, our largest model, i.e., InternVL3.5-241B-A28B, attains state-of-the-art results among open-source MLLMs across general multimodal, reasoning, text, and agentic tasks—narrowing the performance gap with leading commercial models like GPT-5. All models and code are publicly released. > Hatched bars represent closed-source commercial models. We report average scores on a set of multimodal general, reasoning, text, and agentic benchmarks: MMBench v1.1 (en), MMStar,BLINK, HallusionBench, AI2D, OCRBench, MMVet, MME-RealWorld (en), MVBench, VideoMME, MMMU, MathVista, MathVision, MathVerse, DynaMath, WeMath, LogicVista, MATH500, AIME24, AIME25, GPQA, MMLU-Pro, GAOKAO, IFEval, SGP-Bench, VSI-Bench, ERQA, SpaCE-10, and OmniSpatial. In the following table, we provide an overview of the InternVL3.5 series. To maintain consistency with earlier generations, we provide two model formats: the GitHub format, consistent with prior releases, and the HF format, aligned with the official Transformers standard.

번역 대기 중입니다. 원문을 표시합니다.

시각 질의응답license:apache-2.0
컨텍스트 창미공개
입력 → 출력텍스트、이미지 → 텍스트
모델 웨이트미확인
정보 업데이트2026/11/2 02:31:51
OVERVIEW

모델 소개

We introduce *InternVL3.5*, a new family of open-source multimodal models that significantly advances versatility, reasoning capability, and inference efficiency along the InternVL series. A key innovation is the *Cascade Reinforcement Learning (Cascade RL)* framework, which enhances reasoning through a two-stage process: offline RL for stable convergence and online RL for refined alignment. This coarse-to-fine training strategy leads to substantial improvements on downstream reasoning tasks, e.g., MMMU and MathVista. To optimize efficiency, we propose a *Visual Resolution Router (ViR)* that dynamically adjusts the resolution of visual tokens without compromising performance. Coupled with ViR, our Decoupled *Vision-Language Deployment (DvD)* strategy separates the vision encoder and language model across different GPUs, effectively balancing computational load. These contributions collectively enable InternVL3.5 to achieve up to a +16.0\% gain in overall reasoning performance and a 4.05 \\(\times\\) inference speedup compared to its predecessor, i.e., InternVL3. In addition, InternVL3.5 supports novel capabilities such as GUI interaction and embodied agency. Notably, our largest model, i.e., InternVL3.5-241B-A28B, attains state-of-the-art results among open-source MLLMs across general multimodal, reasoning, text, and agentic tasks—narrowing the performance gap with leading commercial models like GPT-5. All models and code are publicly released. > Hatched bars represent closed-source commercial models. We report average scores on a set of multimodal general, reasoning, text, and agentic benchmarks: MMBench v1.1 (en), MMStar,BLINK, HallusionBench, AI2D, OCRBench, MMVet, MME-RealWorld (en), MVBench, VideoMME, MMMU, MathVista, MathVision, MathVerse, DynaMath, WeMath, LogicVista, MATH500, AIME24, AIME25, GPQA, MMLU-Pro, GAOKAO, IFEval, SGP-Bench, VSI-Bench, ERQA, SpaCE-10, and OmniSpatial. In the following table, we provide an overview of the InternVL3.5 series. To maintain consistency with earlier generations, we provide two model formats: the GitHub format, consistent with prior releases, and the HF format, aligned with the official Transformers standard.

주요 특징

特点摘要待依据来源补齐，不凭模型名称推断能力。

한계 및 유의 사항

구체적 한계를 판단할 출처가 부족합니다. 실제 성능, 지연, 데이터 정책, 라이선스와 버전 수명을 확인하세요.

PRICING & COST

출처별 요금 및 비용 분석

요금은 채널, 지역, 설정에 따라 다릅니다. 각 채널 요금은 개별 표시되며 제공사의 통일 공식 요금이 아닙니다.

출처 / 채널입력출력과금 단위수집 시간
확인된 요금이 없습니다. 출처를 확인하세요.

월간 기본 토큰 비용 추정

—

추정치에는 캐시 할인, 구간 요금, 멀티모달 추가 요금, 검색, 세금 및 추가 추론 출력이 포함되지 않습니다.

INDEPENDENT BENCHMARKS

공개 평가 및 실행 설정

暂无已采集的可核验公开评分。模型未上榜不代表能力较低。

등록된 전체 순위 보기 →

자주 묻는 질문

오픈 웨이트면 무료 상업 이용이 가능한가요?

반드시 그렇지는 않습니다. 버전별 라이선스, 상업 이용 조건과 정책을 확인하세요. 자체 배포에는 연산 및 유지 비용도 듭니다.

실시간 결제 요금인가요?

수집 시간이 기록된 요금 스냅샷입니다. 최신 버전, 지역, 할인과 과금 방식은 해당 채널에서 확인하세요.

출처 및 추적 기록

huggingface · OpenGVLab/InternVL3_5-4B-HF ↗2026/11/2 02:31:51
huggingface · enrich:bc7ab22c-a1d9-4c09-aea3-e8fb2a81c9f1 ↗2026/10/29 23:57:53

许可证：apache-2.0。模型介绍与第三方评测均以原来源的具体版本和方法为准。