Distilled DeepSeek into Gemma 4 26B-A4B and 12B
A Reddit user used Unsloth Studio to distill DeepSeek v4 Pro into Gemma 4 26B-A4B and 12B models. The experiment was not very useful but the author reports significant learning.
1 source
Daily brief
Get tomorrow's AI brief in your inbox
More stories today
- User shares trick: ChatGPT creates custom podcasts for car rides
- Hugging Face CEO: Most AI workloads will run on open models
- Enterprises winning with AI agents are limiting agent autonomy
- Sanders to Trump: Have Elon Build a Data Center at Mar-a-Lago
- Offline voice translator runs on-device with Gemma 4 and LiteRT-LM