GPT-5.6 Pricing Explained: Sol, Terra, and Luna API Costs Compared

OpenAI's new GPT-5.6 family does not give you one model — it gives you three, each hitting a different cost-performance point. Understanding GPT-5.6 pricing before the general release lands matters if you are running any non-trivial workload on the API.

What is GPT-5.6, and when does it arrive?

OpenAI announced the GPT-5.6 Sol, Terra, and Luna family on June 26, 2026. As of early July, the models are in a limited preview available to roughly 20 partner organizations, according to VentureBeat — a deliberate rollout shaped in part by US government review. A general API release is expected "in the coming weeks," though no exact date has been confirmed.

The three names map to a tiered strategy: Sol is the flagship (highest capability), Terra is the balanced workhorse (production-ready), and Luna is the lightweight option (fast and cheap). Same model family, different capability-cost trade-offs.

GPT-5.6 Sol, Terra, and Luna API pricing

All three models are billed per million tokens, with separate rates for input and output:

| Model | Input per 1M tokens | Output per 1M tokens | |---|---|---| | Sol (flagship) | $5.00 | $30.00 | | Terra (balanced) | $2.50 | $15.00 | | Luna (lightweight) | $1.00 | $6.00 |

Output tokens dominate real costs — a 1,000-token response from Sol costs $0.03, while the same response from Luna costs $0.006. At scale, that gap compounds fast. For context, according to industry pricing trackers, GPT-4-level reasoning cost around $30 per million tokens in 2023; Luna now delivers comparable throughput at $1-6.

Which GPT-5.6 tier fits which workload?

OpenAI positions the three models for distinct use cases.

Sol is for the hardest tasks: complex multi-step reasoning, extended coding sessions, agent-driven workflows, and security-sensitive applications. If the task demands the best possible answer regardless of cost, Sol is the pick.

Terra targets production environments where you need reliable results at high volume without the flagship overhead. Think customer support pipelines, document processing, or retrieval-augmented generation at scale — workloads where Sol's ceiling is not needed on every call.

Luna is the speed-and-cost optimizer. For quick classifications, short summaries, chat autocomplete, and tasks where the quality delta between "very good" and "best possible" is invisible to the end user, Luna keeps the API bill manageable.

The practical challenge: not every task maps cleanly to one tier. A single session might need Sol for an initial complex analysis, Terra for a follow-up explanation, and Luna for a quick formatting step — and manually switching models mid-session adds real friction.

Sol on Cerebras: 750 tokens per second

For latency-critical applications, OpenAI is deploying GPT-5.6 Sol on Cerebras hardware this July, targeting up to 750 tokens per second. That is aimed at enterprise scenarios where real-time frontier reasoning is the blocking constraint — security monitoring, high-stakes agentic pipelines, or live analysis at the edge. Pricing for the Cerebras-accelerated deployment has not been disclosed separately; expect a premium over standard Sol rates.

What the three-tier model means for your AI spend

GPT-5.6's Sol-Terra-Luna structure makes one thing explicit: "which AI model" is increasingly "which tier of that AI model." Teams that used to choose between providers now also choose between capability levels within a single provider's family.

That tiering logic is the same premise behind multi-model routing tools like ByteChat, where Smart Routing auto-selects the cheapest model capable of answering each specific query — whether that means Luna for a quick lookup or Sol when the task genuinely demands it.

Defaulting to the right tier, rather than the most expensive one by habit, is where most per-token savings actually come from.

Frequently asked questions

How much does GPT-5.6 Sol cost per million tokens?

GPT-5.6 Sol is priced at $5.00 per million input tokens and $30.00 per million output tokens. OpenAI announced this pricing on June 26, 2026; no introductory discount period has been confirmed for the Sol tier.

Is GPT-5.6 available now?

As of early July 2026, GPT-5.6 Sol, Terra, and Luna are in limited preview for approximately 20 partner organizations. General API access is expected within the coming weeks, though OpenAI has not given a specific date.

What is the difference between GPT-5.6 Sol, Terra, and Luna?

Sol is the flagship built for complex reasoning, long coding sessions, and agentic workflows; Terra is the balanced option for high-volume production workloads; Luna is the fastest and cheapest tier for everyday tasks. All three share the same model family, with Sol costing five times more than Luna per input token.