Jono's Corner · 2026-10-07 · notable
Why isn't the industry freaking out about DeepSeek 4.1 Flash? — one dev's month
Developer Jono argues DeepSeek 4.1 Flash is close enough to Claude Opus to be hard to tell apart mid-session, while a task costs about $0.003 instead of $1. The essay drew 350+ points on Hacker News.
A heavy AI-coding user says a cheap open model now feels like a frontier one, and the industry is underrating it.
What is it?
Jono, writing on the blog Jono's Corner, reports on about a month of heavy use of DeepSeek 4.1 Flash across a dozen projects. The essay says that mid-session, without looking at the model name, the author "honestly could not tell you" whether it is DeepSeek or Opus, and it treats DeepSeek 4.1 Flash as a frontier model.
How does it work?
The cost argument rests on the author's own usage: on a $10/month OpenCode Go plan, DeepSeek is "basically unlimited", and a task that would cost $1 on a frontier model costs about $0.003. The essay credits DeepSeek's cache work, saying it shrank the KV cache by roughly 437x compared to the V1 model, for keeping long sessions cheap.
Why does it matter?
The essay reached the Hacker News front page with more than 350 points and 300 comments, a sign that many developers are weighing cheap open-weight models against paid frontier ones. Jono predicts the industry will underrate this shift, and that private, local use will follow ("if your concern is privacy, just wait").
Who is it for?
developers choosing a model for coding agents