Think you need a premium model to run a serious AI agent?
Check again. DeepSeek V4.1 Flash dropped on Sept 10, and you can run it free inside Hermes. 1/ It's a mixture of experts model. 552B parameter backbone, but it only activates ~8B on input and ~16B on output. A big brain
US chip export bans forced China's AI labs to innovate around scarcity.
The result: DeepSeek R1 trained for ~$5.6M matches models US labs spend $100M+ on. The efficiency gap is structural: - US labs optimize for compute abundance — throw more GPUs at every problem - Chinese
DeepSeek v4.1 Flash created an SVG character rig.
Objectively, it met the criteria and the rig worked. I wasn’t blown away by the designs, but maybe that’s beside the point. It actually did what I asked. What do you think?
I’ve been a nomad for the last 3 years (remote work perks), so I built THIS 💪
After spending the last few weeks testing Cline Desktop, finding bugs, talking to beta testers, shaping the launch, and figuring out how we tell the story of the product. The obvious next step was to
You have reached the end of the archive
All of deepseek