GPT-6 Sol leaks, screenshots and release rumors

A screenshot shows GPT-5.6 Sol's description changed from "latest frontier agentic coding model" to "reliable agentic workhorse for everyday tasks" with catalog priority moved from 6 to 4, read as making way for GPT-6 Sol.

A community screenshot shows requests sent as gpt-5.6-sol returning gpt-6-sol in the upstream response metadata, triggering a model mismatch warning, alongside other hints like a third-party platform adding a gpt-6-sol entry and Codex weakening 5.6 Sol's description.

A user says rumors of a GPT-6 Sol release are getting louder and asks OpenAI directly whether it launches today.

xAI published seven hard-work tests in which Grok 4.7 beat GPT-5.6 Sol on coding, circuits, legal work and two others, while an outside lab scored them 46 to 47.

The account scaling01 says GPT-6-Luna and GPT-6-Sol are both coming very soon, without giving a date or evidence.

A user says they are excited for tomorrow, expecting a banked reset, new GPT-6 Sol and Luna models and Opus 5.5, as personal anticipation only.

A ChatGPT Pro user says 12 requests in one Codex session were served as gpt-6-sol, with the server identifying the model as GPT-6 Sol, and links a GitHub issue as evidence.

An imagined weekly lineup runs Grok 4.7 on Monday, Opus 5.5 and GPT-6 Sol plus Luna on Tuesday, Kimi K3.1 on Wednesday and Fable 5.5 on Thursday.

A user criticizes Grok 4.7 as poor, says Anthropic and OpenAI are far ahead, and expects GPT-6 Sol and the new Opus to crush it.

A Japanese roundup says Sol may arrive Tuesday, possibly cheaper and smarter than Astra at around a 6.2 level, and that an unreleased model OpenAI considers AGI, named Bel, reportedly supported its release.

mark_k guesses the rest of the AI week will bring Opus 5.5 on Tuesday and GPT-6 Sol on Thursday, a personal schedule guess.

A roundup of this week's launch rumors: Step 5, Grok 4.7 and MiMo v2.6 are confirmed, GPT 6 Sol and Opus 5.2 are rumors, and Gemini 4.0 is rated very unreliable.