DeepSeek just dropped a new open-source multmodal AI model, Janus-Pro-7B. It is MIT opensource license.
It’s multimodal (can generate images) and beats OpenAI’s DALL-E 3 and Stable Diffusion across GenEval and DPG-Bench benchmarks.
This comes on top of all the R1 hype.
Here is the link to the Deepseek Janus 7B Github.
NEWS: DeepSeek just dropped ANOTHER open-source AI model, Janus-Pro-7B.
It's multimodal (can generate images) and beats OpenAI's DALL-E 3 and Stable Diffusion across GenEval and DPG-Bench benchmarks.
This comes on top of all the R1 hype. The 🐋 is cookin' pic.twitter.com/yCmDQoke0f
— Rowan Cheung (@rowancheung) January 27, 2025
Here is the Huggingface area for DeepSeek Janus Pro 7B.
Janus-Pro is a novel autoregressive framework that unifies multimodal understanding and generation. It addresses the limitations of previous approaches by decoupling visual encoding into separate pathways, while still utilizing a single, unified transformer architecture for processing. The decoupling not only alleviates the conflict between the visual encoder’s roles in understanding and generation, but also enhances the framework’s flexibility. Janus-Pro surpasses previous unified model and matches or exceeds the performance of task-specific models. The simplicity, high flexibility, and effectiveness of Janus-Pro make it a strong candidate for next-generation unified multimodal models.
Model Summary
Janus-Pro is a unified understanding and generation MLLM, which decouples visual encoding for multimodal understanding and generation. Janus-Pro is constructed based on the DeepSeek-LLM-1.5b-base/DeepSeek-LLM-7b-base.
For multimodal understanding, it uses the SigLIP-L as the vision encoder, which supports 384 x 384 image input. For image generation, Janus-Pro uses the tokenizer from here with a downsample rate of 16.
Brian Wang is a Futurist Thought Leader and a popular Science blogger with 1 million readers per month. His blog Nextbigfuture.com is ranked #1 Science News Blog. It covers many disruptive technology and trends including Space, Robotics, Artificial Intelligence, Medicine, Anti-aging Biotechnology, and Nanotechnology.
Known for identifying cutting edge technologies, he is currently a Co-Founder of a startup and fundraiser for high potential early-stage companies. He is the Head of Research for Allocations for deep technology investments and an Angel Investor at Space Angels.
A frequent speaker at corporations, he has been a TEDx speaker, a Singularity University speaker and guest at numerous interviews for radio and podcasts. He is open to public speaking and advising engagements.
One aspect that is generally not well focused on is the decrement of training effort in the context relative to inference. i.e. The ratio of training time to inference time. This is the real metric that should be the focus because it’s going to keep on falling in step changes between the small slides. Training effort is still many, many magnitudes larger and less efficient than the very slow 20W biological form we call our brain…. Were still dealing in matrix math and not full sparsity.
Is Janus Pro 7B free?
Can it be “run on a laptop” like “Using Ollama to Install Deepseek 14B on a Laptop”?
Does Janus then need Internet connection?
If so, it helps explains Nazdaq crash 27 Januar.
What is it like at self driving automation?
Let not be naive of the hour and inning on the world stage — the keyword is “Deep”, as in: DEEP FAKE Seek– Early demonstrations of Western AI Models required much behind-the- scenes human interaction; so it should come as no surprise if Janus Pro 7B is simultaneously taping into and pulling from all the models within its reach. Lets face it, many tech Execs have demonstrated sympathy to the Pan-Das (Huang, et al) and have circumvented international tech embargos by providing unfettered access to the US-based AI platforms, especially since the chip tech has been their firewall and limiting factor. On the flip-side there is much more coming down the pike in near term AI Architecture that will hard code the algorithms and take AI out of its UNIVAC footprint phase.