Here is how to enable a 1M-token context window in Codex for GPT-5.6 Sol.
Even though we have tuned the context limit in Codex to be set optimally when it comes to performance and cost, this is a common ask, so here it is documented.
A larger context window lets Codex retain more code, tool output, and conversation history before summarizing older material. You need a model that supports it. And GPT-5.6 Sol, for example, has a documented 1,050,000-token window.
Open ~/.codex/config.toml and add or update these settings at the top level, before any [section] headers:
```
model = "gpt-5.6-sol"
model_context_window = 1000000
model_auto_compact_token_limit = 900000
```
The first setting selects the model. The second tells Codex to use a one-million-token context budget. The third starts automatic history compaction around 900,000 tokens, leaving some headroom. Restart Codex client and start a new session after saving.
To try the configuration for a single CLI session without changing your defaults:
```
codex -m gpt-5.6-sol \
-c model_context_window=1000000 \
-c model_auto_compact_token_limit=900000
```
Have fun, but also know that we tuned the default carefully!
"You've reached your rate limit. Try again after 7pm."
every. fucking. day.
$200/month - $2,400 a year - for Codex. then i opened the DeepSeek pricing page.
Someone made DeepSeek a native subagent for Codex. Heavy work goes there, light stays on GPT-5.
Same code out the other end. 35x cheaper.
DeepSeek V4 Flash lives on its own API key, not on my Plus quota.
One command:
npx skills add oil-oil/codex-deepseek-subagent -g -y
Restart Codex, tell it "set up DeepSeek as a subagent", done.
First day in a month I'm still working in the evening instead of hitting the wall.
save this before your next rate-limit ↓