@thsottiaux I don’t care about API pricing or “equivalent value.” I care about actual tokens and quality per token. Does $200 Pro give me more, fewer, or the same tokens? GPT-6 Sol currently feels worse than 5.6 Sol. If it’s fewer tokens at lower quality, why not choose Opus 5.5 instead?
@thsottiaux After changing the settings, Sol cannot officially enable 1m context. Only Luna and Terra can enable 1m context. Sol will still revert back to 258k context, so I wrote a skill to solve it. https://t.co/e4fgSKlQqi
@thsottiaux There seems to be a problem with the memory compression of Codex. I gave him a goal mode, and in the middle, I proposed a temporary task, and he started working on it. After completing the temporary task and delivering the materials to me,