Codex has been ridiculously useful lately.
The code review is especially great when I throw Fable 5-generated code at it. Fable 5 writes fast, Codex shows up like the senior engineer asking, “Cool… but what happens when this breaks in production?” 😂
And with Codex’s built-in browser, recon and OSINT workflows are honestly kind of wild. Search, verify, inspect, automate—it goes full beast mode.
Fable 5 writes the code.
Codex reviews it, questions my life choices, and helps clean up the mess.
Pretty solid combo. 😎
If you aren't yet bold enough to install the Codex app, you can stay in the presence of your orange crab and point it at GPT 5.6 Sol. Takes 5 minutes. Kudos to Theo for explaining one of the ways to get this done.
Step 1: Install CLIProxyAPI
Step 2: Connect
Step 3: Define following alias and enjoy claudex
```
alias claudex='CLAUDE_CODE_SUBAGENT_MODEL=gpt-5.6-sol \
CLAUDE_CODE_ALWAYS_ENABLE_EFFORT=1 \
CLAUDE_CODE_MAX_TOOL_USE_CONCURRENCY=3 \
ENABLE_TOOL_SEARCH=false \
claude --model gpt-5.6-sol'
```
If this gets blocked, I owe you a reset.
Luna、Terra、Sol 到底該怎麼選?!
在 Artificial Analysis 的「智慧水準 vs 單任務成本」座標上,Sol 與 Luna 在每種推理強度下都領先 Terra,整個 GPT-5.6 家族都�� GPT-5.5 更划算,唯獨 Terra 沒有甜蜜點:同成本下 Sol 更聰明,同智能下 Luna 更便宜。
想要更高智能就選 Sol,想要更好性價比就選 Luna。
實務怎麼配:
- Luna high:日常寫程式主力
- Luna xhigh:想要更高品質,又不必跳到昂貴模型時
- Sol high:真正需要判斷力的任務
高 effort 的 Luna 幾乎總是勝過 Terra,同等表現卻更便宜;Sol High 以下請直接忽略,改用高 effort 的 Luna;需要 Sol Extra High 時才考慮 Terra Ultra;Sol Ultra 相對 Max 多出的成本通常沒那麼值得。
最後特別要注意 Ultra,他會進到平行Multi-Agent模式,燒錢很兇。務必要求它固定用 1–3 個 agent。