GPT-6 Sol leaks, screenshots and release rumors

A 19-year-old's first vibe-coding project uses Sol to write the app and jev for sub-second responses, with gesture control of a globe; he says four hotel lobbies have paid $11,600.

A long post argues multi-turn agents destabilize after turn 20 and need routing separated from generation, assigning GPT-6 Sol to parallel extraction and Luna to verification while claiming 12.8x speed and 116x lower inference cost.

The author says Opus 5.5 lives up to its reputation with limits barely dropping and high EQ, while GPT-6 Sol is cheap but underpowered and takes over tasks previously handled by GPT-5.6 Luna, and questions whether OpenAI plans to offer the GPT-6 series in web ChatGPT.

The author lists this week's new models, including Grok 4.7, Opus 5.5, GPT-6 Sol and Luna, and jokes that next month everyone may still just use Jev.

A recommended cost-saving workflow routes most coding to Luna medium, judgment to Jev, and only complex tasks to GPT-6 Sol, sharply cutting token use.

The episode covers the boom in personal agents like Instinct, Grok Bot and Muse, the Jev classification model, and early impressions of Opus 5.5 and GPT-6 Sol.

Third-party platform Manyways announced support for the new GPT-6 Sol and Luna models, letting users run their own prompts in Dispatch to compare answers.

A weekly roundup says GPT-6 Sol and Terra are roughly 50% cheaper and faster, Opus 5.5 is about 40% cheaper and noticeably better at writing, and also mentions JEV, Muse and Grok 4.7.

a16z's Levie says Opus 5.5 and GPT-6 Sol/Luna cut prices on the same day, and that falling cost per task will open up large-scale agent use cases like a Jevons paradox.

A user jokes that Jev has already appeared while GPT-6 Sol is still missing, reflecting the community's wait for Sol's release.

A comparison chart claims the decision model Jev is about 238x cheaper than GPT-6 Astra, has a 0% error rate on structured output and tool calls, and places Sol/Opus 5 in the 73–74% accuracy tier.