GPT-6 Sol leaks, screenshots and release rumors

MBZUAI's ThinkBooster orchestrates GPT-6 Sol, Luna, Opus 5.5 and Grok 4.7 in tree search, lifting HumanEval+ by 9.5% while cutting compute by 68%.

A DevDay wish list covering Astra 6.1 fixes, Voice mode 2, Ultrafast versions of GPT-6 Sol and Luna, a hardware product and more.

An opinion post says many Codex users moved to Claude over Opus 5.5 but haven't fully left, waiting for Tuesday's DevDay to see if GPT-6 Sol can win them back.

The API value-comparison project updated quota data including ChatGPT Plus × GPT-6 Sol and added new leaderboards and chart interactions.

A Mac app developer says GPT-6 Sol is slow, gets stuck in local optimization loops, produces sloppy UIs and forgets instructions, with low cost as its only advantage.

A Polish creator recaps an in-person meetup and notes the blog already covers the AI Act, Opus 5.5, a ranking of Chinese models and the latest on GPT-6 Sol/Luna.

A user who says they use AI every day calls GPT-6 Sol terrible: it loses the thread of conversations and writes worse than any recent GPT model, feeling like a rush job.

A $200/month user says three days of GPT-6 Sol xhigh used 60% of their quota for sloppy output, and threatens to move to Claude Opus 5.5 if DevDay disappoints.

The poster says GPT-6 Sol has been frustrating since launch and wants DevDay to bring a model that makes Opus 5.5 look outdated.

A user rounds up more than ten model releases this week, including GPT-6 Sol, GPT-6 Luna, Opus 5.5 and Grok 4.7, and teases Sonnet 5.5.

The author argues Anthropic optimizes for the smartest model while OpenAI's GPT-6 Sol targets the most-used one, near Fable 5.1 level at a tenth the cost and faster.

The author says Anthropic and OpenAI shipped Opus 5.5 and GPT-6 Sol within minutes of each other on Tuesday and scored seven of the best matchups.