Deepseek-V4-Flash Leads Open Model Usage on OpenCode With 1.27T Tokens

DT

Daniel T. Vela@danieltvelaTECH

@omarsar0 @dair_ai Don’t hesitate to try a small, fast and cheap model as orchestrator. It’s surprasingly very good. DeepSeek-V4-Flash it’s better than the one I use. Let subagents do the deep work (Oracle, Librarian). The Inverted Rule of muti-agents harnesses. https://x.com/danieltvela/status/2074554555728019726?s=20