GPT-6 Sol leaks, screenshots and release rumors

A report says GPT-6 Sol and Luna come with outside testing rules, leading enterprise buyers to ask vendors for audit evidence beyond benchmark scores.

A poster says GPT-6 Sol gains 2 points on the Artificial Analysis Coding Agent Index at half the cost per task, while Luna drops 2 points and falls behind Gemini 3.8 Flash.

Vercel says GPT-6 Sol and Luna are now available on AI Gateway at half the price of GPT-5.6, building on Astra's advancements.

Given the same prompt and task, the author says Claude Opus 5.5 has clearly better taste and design than GPT-6 Sol, which improves on GPT 5.6 Sol but falls well short of Opus 5.5.

A post claims Opus 5.5 already beat GPT-6 Astra and that Anthropic plans to release Sonnet 5.5 and Haiku 5.5 within weeks to counter GPT-6 Sol and Luna.

A user shares an openai.com link titled "introducing gpt-6 sol and luna" with the caption "Models for the people," suggesting an official launch page is live.

A post claims the halved pricing for GPT-6 Sol and Luna was decided at the last minute before launch, an unconfirmed rumor.

A video post puts Opus 5.5 and the rumored GPT-6 Sol side by side for a direct comparison.

A Spanish-speaking user says Terra no longer offers good value and Luna or Sol are the better pick, now upgraded to GPT-6 and cheaper, used under OpenCode2 and Codex.

The author shares a cycling video made with GPT-6 Sol and complains the motion is too fast and the result is hard to describe.

A Spanish-language post notes that Anthropic launched Opus 5.5 and OpenAI responded hours later with GPT-6 Sol and Luna, putting the two labs' releases almost back to back.

GitHub's changelog says GPT-6 Sol is the balanced model for interactive and agentic coding, while GPT-6 Luna is a lightweight, low-cost option for smaller tasks, and both are now generally available in Copilot.