Two ways to keep using Codex after usage limits tighten
With Codex usage cut, the author suggests two options: use Codex Router to add cheaper models alongside Sol and Astra instead of replacing them, or have Codex drive a browser to use GPT-6 Pro in ChatGPT.
Codex usage is so gone now. 2 ways we can keep going. 1. Add a cheap model without losing your GPT ones. DeepSeek's own setup points your whole Codex at their API, so Sol and Astra disappear. Codex Router adds to your model picker instead of replacing it, so DeepSeek V4.1 Flash and Muse Spark 1.3 just sit next to them. $10 a month on OpenCode Go is 26,000 Flash requests every 5 hours. 2. Let Codex use Web ChatGPT for you. Run GPT-5.6 Luna, the cheapest model, on the lowest thinking setting, and let it drive a browser. It only clicks and types, so it burns almost nothing, and the answer comes back from GPT-6 Pro. Pro isn't in Codex at all and it's a different usage than codex but it only exists in chat. Keep on going everyone!