Improving GPT-5.6 Sol in ChatGPT—and expanding access for free users

Source: OpenAI Date: 2026-08-06 URL: https://openai.com/index/improving-gpt-5-6-sol-in-chatgpt

Summary

OpenAI tightened GPT-5.6 Sol’s behavior for Plus/Pro subscribers — more focused answers, better factual reliability, tighter formatting, and a willingness to correct rather than just agree with the user — and added a reasoning-effort slider. Separately, Free-tier users get unlimited text chats on GPT-5.6 Luna (becoming their default model) plus a “Think” button for on-demand higher reasoning, rolling out over the following week.

Implications

Feeds the model-landscape and cost-economy threads: OpenAI is segmenting its lineup by tier (Sol for paying users, Luna as the free default) rather than shipping one model to everyone, and is competing on free-tier generosity (unlimited chats) the same way Google and others have been. Also a small trust-adjacent note — reducing sycophantic agreement in favor of “helpful corrections” is a direct response to the criticism that chat models over-validate users. Worth watching whether the reasoning-effort slider pattern spreads to other vendors’ consumer surfaces.