OpenAI has begun a limited preview of GPT-5.6, splitting its next-generation model into three named tiers — Sol, Terra, and Luna — in a structural shift that introduces tiered pricing and two new reasoning modes, MarkTechPost reports.
Sol is the flagship, targeting the highest-intelligence workloads with state-of-the-art performance on benchmarks like Terminal-Bench 2.1. Terra matches GPT-5.5 capability at roughly half the cost, positioned as the production workhorse. Luna is the low-cost option — strong capability at OpenAI's lowest price point. The numbering now marks the generation, while the names mark durable capability tiers that can advance on independent schedules.
The two new reasoning modes are "max" — which gives Sol more time to reason deeply on a single chain — and "ultra," which coordinates multiple subagents to split and accelerate complex work. OpenAI shared the models with the US government first, and broader access via ChatGPT, Codex, and the API is planned in the coming weeks.
Analysis
The tiered naming — Sol, Terra, Luna — is not an aesthetic choice. It is a pricing architecture. By decoupling the generation number from the capability tier, OpenAI is building the infrastructure for differential monetisation: Sol for the customers who will pay anything for state-of-the-art, Terra for the volume enterprise market that cares about cost-per-token, and Luna for the free-tier and API price-war segment. This is the same playbook every cloud provider runs — premium, standard, economy — and it signals that OpenAI is preparing for a world where inference is a commodity business with razor-thin margins at the bottom and fat margins at the top. The "ultra" subagent mode is the real pricing innovation. It doesn't just consume more tokens — it consumes them in parallel across multiple agents, multiplying the per-query cost. OpenAI is building a reasoning product where the customer's bill scales with problem complexity, not just token count. That is not a feature. It is a revenue model.
