After pushing 2B tokens through DeepSeek V4.1 Flash, dev cuts AI bill from $300 to $30/month

gaganghotra_ · x · 2026-09-18

Developer @btsouth reports running 2+ billion tokens through DeepSeek V4.1 Flash in 4 days across his agent workflows. His monthly AI spend dropped from $300 — with constant usage-anxiety — to $30, letting him run agents all day without rationing tokens. He claims the model is remarkably close to frontier quality for his work, with no situation in 4 days of heavy use where he needed something better, and argues this isn't just a honeymoon period. His new stack: DeepSeek + OpenCode + Ollama Cloud, which he says has reset his expectations of what AI should cost. Note: the model name comes from a personal post and is not yet officially confirmed.

Original post →

More from coding & agent

coding & agent channel →