谷歌发布第三代开源大模型Gemma 3
SOURCE / DeepMind-官方网站 · gemma 3
原文
gemma 3
完整原文
Gemma 3A family of lightweight models with multimodal understanding and unparalleled multilingual capabilities for more intelligent applications.Explore Gemma 3 Read docs View tech report Gemma 3 is the most capable model that can run on a single GPU or TPU. Efficient on workstations, laptops, and even smartphones, allowing developers to build responsible AI applications at scale. Capabilities Models Performance Download CapabilitiesHandle complex tasksGemma 3's 128K-token context window lets your applications process and understand vast amounts of information, enabling more sophisticated AI features.Multilingual communicationUnparalleled multilingual capabilities let you communicate effortlessly across countries and cultures. Develop applications that reach a global audience, with support for over 140 languages.Multimodal understandingEasily build applications that analyze images, text, and video opening up new possibilities for interactive and intelligent applications.Model variants270MCompact model designed for both task-specific fine-tuning and strong instruction-following.1BLightweight text model, ideal for small applications.4BBalanced for performance and flexibility, with multimodal support.12BStrong language capabilities, designed for complex tasks.27BEnhanced understanding, great for sophisticated applications.PerformanceSlide 1 of 9MMLU-ProThe MMLU benchmark is a test that measures the breadth of knowledge and problem-solving ability acquired by large language models during pretraining.LiveCodeBenchAssesses code generation capabilities on real-world coding problems from platforms like LeetCode and Codeforces.Bird-SQLTests a model's ability to translate natural language questions into complex SQL queries across various domains.GPQA DiamondChallenges models with difficult questions written by Ph.D. holders in biology, physics, and chemistry.SimpleQAEvaluates a model's ability to answer simple, factual questions with short phrases.FACTS GroundingEvaluates if LLM responses are factually accurate and detailed enough, based on given input documents.MATHMATH evaluates a language model's ability to solve complex mathematical word problems, requiring reasoning, multi-step problem-solving, and the understanding of mathematical concepts.HiddenMathAn internal holdout set of competition math problems.MMMUEvaluates multimodal understanding and reasoning across various disciplines requiring college-level knowledge. Gemma Quantization-Aware Training (QAT)Gemma QAT dramatically reduces memory requirements while maintaining high quality. This lets you run powerful models like Gemma 3 27B locally on consumer-grade GPUs like an NVIDIA RTX 3090. Read more What’s new in Gemma 3Slide 1 of 5Welcome to Gemma 3Discover the latest advancements in Gemma, Google's family of lightweight, state-of-the-art open models.Deepdive into Gemma 3Hear how the Gemma research team unveil the architecture, design principles, and innovations behind Google's family of lightweight, state-of-the-art open models.Building intelligent agents with Gemma 3Explore the development of intelligent agents using Gemma models, with core components that facilitate agent creation, including capabilities for function calling, planning, and reasoning.Gemma 3 architecture and designDiscover how with Gemma 3, we have tried to push many of the limits of what makes a model usable and practical.A truly multilingual Gemma 3Building multilingual AI applications is crucial for reaching global audiences, and varied language proficiency remains a top developer priority. Download Gemma 3 Hugging Face Download Ollama Download Kaggle Download
归纳
谷歌近日推出第三代开源大语言模型Gemma 3，参数规模包括2B、5B、14B等多个版本。该模型采用兼容性更强的新许可证，允许商业使用和修改，但用户需严格遵守许可条款及当地法律法规。Gemma 3在多项基准测试中较前代显著提升，支持多语言和长上下文，并可在个人电脑等本地设备运行。模型权重已在Hugging Face等平台公开发布，方便开发者用于研究、产品开发及教育用途。此次发布进一步完善了开源AI生态，使用过程中须关注数据隐私、模型安全及合规性等法律问题。
点评
Gemma 3开源商业许可与中国生成式AI备案及内容审查制度存在直接冲突，用户需确保本地部署和修改符合算法透明度与数据安全义务。
法律视角点评
AI 生成 · 人工审核核心关切
Gemma 3开源商业许可与中国生成式AI备案及内容审查制度存在直接冲突，用户需确保本地部署和修改符合算法透明度与数据安全义务。
实务启示
中国法律人应协助企业将开源许可条款与本土AI监管要求对比审查，重点厘清修改后的模型是否触发备案及内容管理责任。