原创童欧巴欧巴聊A12025-01-28090336收录于话题大家好，我是欧巴。今天是除夕，本想睡个懒觉，梦里还回味着DeepSeek AppStore登顶的欣喜若狂。突然被我妈叫醒贴对联，看到手机屏幕上不断翻出来新消息：DeepSeek Janus-Pro开源然后就有了这篇文章，让我们一起感受下年关前的惊喜。deepseek-ai/Janus-Pro-7B like 788Follow DeepSeek 14.2kAny-to-AnyTransformersO PyTorchmulti_modalitymuiltimodaltext-to-imageModel card恒Files and versionsCommunity 64Janus-ProJaus-Pro是一个多模态模型，搜了一下，名称来源于古罗马神话种中的门神雅努斯”：拥有两张面孔，一张看向过去，另一张面向味来。象对时间与空间的掌控，以及在不同状态间的转换能力。刚好对应归us-Pro的两种能力，既可以进行视觉理解，也能进行图像生成。相关的发布地址如下：https://github.com/deepseek-ai/Janus.https://huggingface.co/deepseek-ai/Janus-Pro-1B.https://huggingface.co/deepseek-ai/Janus-Pro-7B同时还有一份技术报告，回复Janus获取中英对照版本。官方给出了一些效果图，感受一下：JanusJanus-Pro-7BJanusJanus-Pro-7BJanusJanus-Pro-7BThe face of a beautiful girl.A steaming cup of coffee on a wooden table.A glass of red wine on a reflective surface.0HelloA minimalist photo of an orange tangerine A clear image of a blackboard with a clean,Capture a close-up shot of a vibrant sunflowerwith a green stem and leaves,symbolizing dark green surface and the word'Hello'written in full bloom,with a honeybee perched on itsprosperity,sitting on a red silk cloth during precisely and legibly in the center with bold,petals,its delicate wings catching the sunlight.Chinese New Year.white chalk letters.Figure 2 Comparison of text-to-image generation between Janus-Pro and its predecessor,Janus.Janus-Pro delivers more stable outputs for short prompts,with improved visual quality,richer details,and the added ability to generate simple text.The image resolution is 384 x 384.Best viewed on screen.我的感受是发布即为第一梯队，虽然没有Midjourney出的图精致。还是那句话：让子弹飞一会。继续看数据，在多个基准测试中，Janus-Pro已经超越了OpenAl的DALL-E3和SD3-Medium等一众模型，直接霸榜。Janus-Pro-7B★TokenFlow-XLJanus-Pro-1B64★LLaVA-v1.5-7B*●VILA-U60Emu3-Chat48.●LaVA-V1.5-Phi-1.5*4052Show-oJanus-Pro Family (Unified Model)LLaVA Family (Understanding Only)261012LLM Parameters (Billions)GenEvalDPG-Bench(a)Average performance on four multimodal understand-(b)Performance on instruction-following benchmarks foring benchmarks.text-to-image generation.除了画图之外，Janus-Pro还具备图像识别能幼，也就是能读懂图片。比如让他详细描述如下的场景。