GPT-6 Sol leaks, screenshots and release rumors

The poster notes Codex users stopped asking for weekly limit resets, guessing GPT-6 Sol is cheap enough that quotas last or everyone moved to Opus 5.5.

Third-party platform Dot says it now hosts the full GPT-6 family (Astra/Sol/Luna) with per-million-token pricing it claims is 10% below standard OpenRouter rates.

On the same drawing task Opus 5.5 took 1h 23m and cost $6.31, while GPT-6 Sol took 26 minutes and $14.97 but produced a worse result.

A user says Sonnet 5 is routing to 5.5, the team confirmed an upcoming release, and early output looks much better than GPT-6 Sol, with a comparison video attached.

Caffeine now lets users pick which frontier model builds their app, with GPT-6 Astra and GPT-6 Sol both listed, and supports switching to a cheaper model mid-build.

Given the same prompt, Opus 5.5, GPT-6 Sol, GPT-6 Luna and Grok 4.7 each built a launch site for a fictional handheld console, with time and cost compared.

The poster speculates that GPT-6 Sol is not yet in ChatGPT because OpenAI plans to announce a merger of the Chat and Work entry points at DevDay.

The poster says Sol's usage pricing is very competitive, but Opus 5.5 is a tier above on capability, and their ongoing Sol tests are not looking good.

A side-by-side comparison finds Opus 5.5 stronger at coding and design as a flagship model, while GPT-6 Sol is a cheaper, faster tier with good value but not top-end performance.

A benchmark chart claims Opus 5.5 outperforms Fable 5 and GPT-6 Sol on vision tasks but falls short of GPT-6 Astra, with cost and reasoning gains listed.

In tests including GPT-6 Sol and Astra, Sol matched Opus's score, finished faster, and cost about a quarter as much per successful task.

Sol [xhigh] wrote a single-launch Kimi-Linear Decode kernel for the RTX PRO 6000 that passed all six checks but ran at 0.96x the optimized PyTorch baseline.