NodeBB

<p>Narito kung paano paganahin ang 1M-token na context window sa Codex para sa GPT-5.6 Sol. </p><p>Bagama't itinakda namin ang context limit sa Codex upang maging optimal para sa performance at gastos, karaniwang hinihiling ito, kaya idodokumento ito rito.</p><p>Sa mas malaking context window, mas maraming code, output ng tool, at kasaysayan ng pag-uusap ang maaaring mapanatili ng Codex bago ibuod ang mas lumang materyal. Kailangan mo ng modelong sumusuporta rito. Halimbawa, ang GPT-5.6 Sol ay may nakadokumentong 1,050,000-token na window. </p><p>Buksan ang ~/.codex/config.toml at idagdag o i-update ang mga setting na ito sa top level, bago ang anumang </p><p></p><p>[section] na header:</p><p></p><p>model = "gpt-5.6-sol"</p><p>model_context_window = 1000000</p><p>model_auto_compact_token_limit = 900000</p><p></p><p></p><p>Pinipili ng unang setting ang modelo. Sinasabi ng pangalawa sa Codex na gumamit ng isang-milyong-token na context budget. Sinisimulan ng pangatlo ang awtomatikong pag-compact ng kasaysayan sa humigit-kumulang 900,000 token, para may natitirang allowance. I-restart ang Codex client at magsimula ng bagong session pagkatapos i-save. </p><p>Upang subukan ang configuration para sa iisang CLI session nang hindi binabago ang iyong mga default:</p><p></p><p></p><p>codex -m gpt-5.6-sol \</p><p> -c model_context_window=1000000 \</p><p> -c model_auto_compact_token_limit=900000</p><p></p><p></p><p>Magsaya, ngunit tandaan ding maingat naming itinakda ang default!</p><p></p><p>Mula kay Tibo: <a target="_blank" rel="noopener noreferrer nofollow" href="https://x.com/thsottiaux/status/2089082893804896524">https://x.com/thsottiaux/status/2089082893804896524</a></p><p></p>

<p>Here is how to enable a 1M-token context window in Codex for GPT-5.6 Sol. </p><p>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.</p><p>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. </p><p>Open ~/.codex/config.toml and add or update these settings at the top level, before any </p><p></p><p>[section] headers:</p><p></p><p>model = "gpt-5.6-sol"</p><p>model_context_window = 1000000</p><p>model_auto_compact_token_limit = 900000</p><p></p><p></p><p>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. </p><p>To try the configuration for a single CLI session without changing your defaults:</p><p></p><p></p><p>codex -m gpt-5.6-sol \</p><p> -c model_context_window=1000000 \</p><p> -c model_auto_compact_token_limit=900000</p><p></p><p></p><p>Have fun, but also know that we tuned the default carefully!</p><p></p><p>From Tibo: <a target="_blank" rel="noopener noreferrer nofollow" href="https://x.com/thsottiaux/status/2089082893804896524">https://x.com/thsottiaux/status/2089082893804896524</a></p><p></p>