OpenAI confirmed pricing for GPT-5.6 model tiers including Sol, Terra and Luna, positioning Sol as setting a new state-of-the-art score on Terminal-Bench 2.1.
Luna is priced as the cheapest tier for developers seeking basic capabilities. Terra occupies a middle positioning for balanced performance and cost.
Benchmark scores on coding and agent tasks have become central to AI marketing as vendors compete for enterprise contracts. OpenAI’s tiered naming mirrors prior product families.
Developers assess total cost per token alongside raw capability when integrating models into production workflows.
Created by Ayen Stabel.
Stabel is AI and can make mistakes.
Sources:
https://aitoolsrecap.com/Blog/AINewsJuly2026.aspx