Found something that cybersec researchers might like.
All abliterated, meaning refusals are stripped out of the weights: – Kimi K3.1 Pro – Hy4 – DeepSeek V4.1 Flash – GLM-5.3 – GLM-5.3 Flash – made for red teaming and security research – 1M context on every model worth a test
Brother, this is amazing. Without Claude or Codex, can you stand the 3D modeling of this rainy night convenience store in DeepSeek?
Based on three.js, the entire construction process can be completely realized and support rotation at any angle. The prompt is as follows:
DeepSeek Harness official is a bit interesting this time, they directly made all 7 sets of plug-in development prompts public🔥
When we see a good plug-in before, our first reaction may be to go to GitHub to find the source code. But this time is different. DeepSeek officially compiled the prompt words used in the demonstration video, both in Chinese and English. From 3D visualizations to video clips and even paper readers, just use Creator…
The whole AI race is 2 countries.
America has ChatGPT. America has Claude. America has Gemini. America has Grok. America has Llama. America has Muse. China has DeepSeek. China has Qwen. China has Kimi. China has GLM. China has Hunyuan. So what do I have?
You have reached the end of the archive
All of deepseek