You can run Opus 4.6 Max for free now.
Not the API.
Qwen3.8-27B on your machine, then pick it inside Codex.
SWE-bench Pro: 61.7 vs Opus 53.4
QwenSWEBench: 79.0 vs 63.8
LiveCodeBench: 90.3 vs 88.8
Pull it in Ollama or llama.cpp.
OpenCodex puts the local model in the Codex picker.
https://t.co/pQ1DG5yqBR
This is Insane
China has more options than america's model
GLM-5.3 release
cheaper than kimi
low size
I cant imagine glm scale more
and you can use live in opencodex
via zcode, opencode-go and any supported
providers
Super Grok’s price needs to be lowered to less than half.
In fact, the Super Grok Heavy $300 plan
only becomes cost-effective
now that Cursor is offering $200.
They give $300 worth per week,
and even with a reset + 0.5 weeks, you can only get about $1,700 worth.
With Cursor, they give an extra $1,400 worth,
so it just barely exceeds $3,000.
OpenAI or Claude, on the other hand,
provide a full $6,000 worth
for just $200.