Future Observation News
Text size

Text size is saved only in this browser.

Updated Trending on XNew GPT-6 models

GPT-6 Sol and Luna arrive as Tibo announces lower prices and a reset

Imagined developer preparing to begin new work
AI-generated editorial illustration, not an actual product screen or documentary photograph. Mirai Kansoku / AI-generated

At 3:23 a.m. Japan time on September 23, Tibo announced GPT-6 Sol and Luna on X, linking OpenAI’s release. This follows his reset teaser and comments about efficiency and intelligence.

OpenAI lists API prices per million tokens at $2 input/$10 output for Sol and $0.10/$0.50 for Luna. Rollout is gradual in ChatGPT Work and Codex; ordinary Chat is not included yet.

Tibo also said a banked reset was being added for Plus, Pro and Business users; we have not verified completion for every account. The follow-up showed over 4,500 likes and about 580 reposts at our check.

What led to the statement

OpenAI describes extending Astra-related training methods and serving efficiencies to cheaper models. That provides the context for Tibo’s earlier emphasis on efficiency and intelligence for everyone.

Our analysis

Lower prices can create entirely new tasks

The larger effect may be on work previously considered too small to delegate: daily document sorting, short writing checks and tiny prototypes. Lower costs could make these continuous habits instead of weekend chores.

This raises the value of dividing work well. Sorting suitable for a cheaper model can be separated from decisions needing deeper judgment, helping small businesses use a limited budget more effectively.

How could everyday life change?

The following is a possible future based on this news.

Review overnight progress in the morning

A future independent developer might queue work before bed, letting AI tackle it as capacity becomes available. Mornings would begin with reviewing completed screens and unresolved questions.

That would require reliable saved progress and stopping rules for failures. The teaser does not deliver this scenario, but planning when AI works could become part of planning the work itself.