GPT-6 Sol leaks, screenshots and release rumors

A short post says a claim aligns with what was seen with GPT-6 Sol and Luna, and that the most important aspect is how much their pricing dropped.

A short speculative post suggests that GPT-6 Sol is actually Terra under a new name.

On the same neon pinball prompt, GPT-6 Luna finished in 6 minutes for about $0.08 and Sol in 17 minutes for about $1.20, while Grok 4.7 took over two hours and cost more.

A security researcher says GPT-6 Sol feels like a significant downgrade from GPT-5.6 Sol for security research and asks whether others feel the same.

A Japanese-language review compares GPT-6 Sol and Luna with Opus 5.5, finding Sol a solid Codex main agent but behind Astra on code review, and Luna cheap, fast and the best current implementation model.

A long review lists Opus 5.5's advantages in finished output quality, coding benchmark scores and cost, and notes most people preferred Opus over GPT-6 Sol in same-prompt comparisons.

A long post breaks down the Rails agent evals, noting Luna is efficient on atomic tasks but drops to zero completion once tasks become full feature tickets, with a comparison to Sol.

After comparing Opus 5.5 and GPT-6 Sol on different tasks, a user says Claude's output is more readable and seems smarter, and is considering making it the primary model again, though it still adds too many code comments.

After seeing the Sol leaks and benchmarks, the poster feels it is more about efficiency than a leap in intelligence, with impressive cost but short of expectations.

Magnific gave Claude Opus 5.5 and GPT-6 Sol the same scene brief, generated their takes with Seedance 2.5, and shared the prompt and timings.

The author says Sol is very fast with less waiting, making game prototyping in a Codex workflow smoother with Combos CLI.

The author recommends Opus 5.5 first, calls GPT-6 Sol a smarter 5.6 Sol, and says Luna is cheap enough for near-unlimited use on non-coding tasks.