view reply Deepseek flash is dramatically less expensive to hose than any of the glm or Kimi models and unlike the other providers deepseek gets a ton of value from their users data. Not sure if there’s anything all that mysterious here.
deepseek-ai/DeepSeek-V4-Flash-0731 Text Generation • 304B • Updated 25 days ago • 3.86M • • 3.73k
Running 29 Soofi S — Sovereign German-English Foundation Model 🚀 29 Soofi S 30B-A3B pretraining report project page