GPT-5.6 Sol launches with coding benchmark leads over Fable 5

GPT-5.6 Sol (max) in Codex leads the Coding Agent Index and ties Grok 4.5 on SWE-Atlas-QnA, with lower cost per task than Claude Fable 5. Early testers describe Sol as similar in ability to Fable 5 but faster and more collaborative.