Dsh-game-studio by FiniYang: an open source DeepSeek Harness plugin that turns one sentence into a playable 2D or 3D voxel game in the chat sidebar.
The agent playtests its own build (frame rate, errors, screenshots), then exports one offline HTML file.
Just crossed 20,000+ model torrents 🧲
Pick, download, and preserve models such as Kimi K3.1 Pro, DeepSeek V4.1 Flash 1017, and more. Pirate Face salutes its community, the peer-to-peer distribution layer for sovereign AI. Those who control the weights control the future ↓
I also used Deepseek V4.1 flash to make a short video of a pelican riding a bicycle in Hermes Agent for fun.
Tip: Generate an animation of a pelican riding a bicycle. Try to make it as interesting as possible. The length of the animation is about 20 seconds. Pay attention to details and quality.
"Open model" means 3 different things
1. Open weights: download them, under the maker's terms (Llama 4) 2. Open weights + MIT/Apache: use, change, sell (gpt-oss, DeepSeek-R1) 3. Fully open: weights, training code and data (OLMo) Most "open-source" LLMs are #1 or #2.
You have reached the end of the archive
All of deepseek