Use Kaggle's free GPU to run the inference model DeepSeek with one click without queuing, with a 30-hour quota per week.
Another post where DeepSeek-v4.1-flash absolutely cooked.
I have been mostly building modern style games with it, so wanted to try something old for a change. Asked DeepSeek-v4.1-flash to build an old-school console RPG, like a tiny SNES/PS1 D&D toy game (plus some more
Union Alpha is a TERRIBLE model
It turns out it's just a router between models it's based on Llama, GLM, DeepSeek, and GPT no new models or technologies moreover, all requests go to third-party resources therefore, ZDR may also be a lie disappointment
We release Needle 3: A Sliceable 8-29MB automation foundation model that can match DeepSeek V4 Flash.
One set of weights, every depth from 2 to 20 layers a model of its own, 25-121M parameters at CQ2-bit, built on our Simple Attention Networks and running locally at up to 4k
You have reached the end of the archive
All of deepseek