OpenAI plans to roll out its GPT-5.6 lineup next Tuesday, likely releasing the standard, Mini and Pro variants at once—though last-minute staggering across ChatGPT, API and Pro channels can’t be ruled out. A GPT-5.6 Pro build has already surfaced for some subscribers, and early tests show stronger outputs compared with GPT-5.5. The big upgrade centers on a context window that jumps from 1 million tokens to about 1.5 million, plus faster, more accurate long-range coding via Codex. The move targets Anthropic’s Mythos series, where developers say GPT-5.6 outperforms Claude Fable 5 on agentic coding tasks. OpenAI’s token pricing already runs at roughly half Anthropic’s rates, and insiders expect further cuts to spark a price war. This push comes as Claude Fable 5 faces U.S. regulatory scrutiny that could limit its reach—leaving a window for OpenAI to grab more market share. Behind the scenes, OpenAI is also testing GPT-Bidi-1, a next-gen bidirectional voice model that can listen and speak simultaneously, handle interruptions and tweak its tone in real time. Inside ChatGPT it’ll appear alongside the current Advanced Voice Mode, with High, Medium and Instant quality tiers. A new draggable voice bubble spotted in the interface hints at the user experience changes to come.