OpenAI announced GPT-5.6 on June 26, 2026, two months after GPT-5.5, and it arrives as three separate models rather than one: Sol, Terra, and Luna. The lineup is in limited preview inside Codex and the API for a select group of trusted partners and organizations, with broader availability planned for the coming weeks. The headline additions for anyone who builds with these models are a new "max" reasoning effort on Sol and an "ultra" mode that spins up subagents to handle work beyond what a single agent can do.
What happened
GPT-5.6 is OpenAI's next frontier generation, and it ships with a renamed structure. In OpenAI's words, "the number identifies a model's generation, while Sol, Terra, and Luna identify durable capability tiers that can advance on their own cadence." In plain terms: the 5.6 is the generation, and the three names are quality-and-cost tiers that OpenAI can update independently instead of bumping a single monolithic model.
OpenAI is rolling it out as a staggered preview, citing coordination with the U.S. government before wider release. The first improvements OpenAI calls out are in coding, biology, and cybersecurity, with the company noting the cybersecurity work shipped "with our most robust safeguards to date."
The three models: Sol, Terra, Luna
Instead of guessing which model fits a task, GPT-5.6 makes the tradeoff explicit. Sol is the flagship for the hardest reasoning, Terra is the balanced everyday workhorse, and Luna is the fast, cheap option for high-volume jobs.
| Model | Tier | Best for | Input / Output (per 1M tokens) |
|---|---|---|---|
| Sol | Flagship | Deep reasoning, complex coding, agentic multi-step work | $5 / $30 |
| Terra | Balanced | Everyday work tasks, drafting, mid-complexity coding | $2.50 / $15 |
| Luna | Fast / affordable | High-volume, latency-sensitive, simple calls | $1 / $6 |
Cache writes are billed at 1.25x the uncached input rate, and cache reads get a 90% discount, so prompt-heavy and repeated-context workflows get materially cheaper if you structure them around caching.
What is new: max reasoning and ultra mode
Two features matter most if you build agents or coding tools:
- Max reasoning effort (Sol). A new effort level that gives Sol the most time to reason deeply on a single hard problem, above the existing effort tiers. Use it when correctness on a gnarly bug or a multi-constraint plan matters more than speed.
- Ultra mode. Rather than thinking harder in one pass, ultra leverages subagents to break a task apart and work beyond what a single agent can do. This is the model orchestrating helpers internally, aimed at long, multi-part jobs like refactoring across a codebase or running a research-and-synthesize loop end to end.
The pattern here is the same one we keep seeing across the field: deeper single-thread reasoning for hard problems, plus an orchestration mode for problems that are large rather than just hard.
Why it matters for creators and builders
The tiered split is the practical story. A creator wiring an AI feature into an app no longer has to overpay Sol prices for a task Luna can do, or settle for a weaker model on the one step that actually needs Sol. You route per step: Luna for bulk classification and quick replies, Terra for everyday generation, Sol with max reasoning for the few calls where being right is worth $30 per million output tokens.
Ultra mode is the piece worth watching for anyone building coding agents or multi-step content pipelines. Subagent orchestration handled by the model removes some of the scaffolding you would otherwise build yourself, though in a limited preview the real test is reliability on long jobs, not the demo.
What this enables
A concrete way to put GPT-5.6 to work once you have access:
- Map your pipeline to tiers. List every model call in your app and tag each one Luna, Terra, or Sol based on how much reasoning it genuinely needs. Most pipelines are overpaying on at least one step.
- Reserve max reasoning for verification. Keep generation on Terra or Luna, then run a single Sol max-reasoning pass to check or repair the output. You pay flagship prices once, not on every token.
- Pilot ultra on one long task. Pick a job that currently needs your own orchestration code (a multi-file refactor, a research brief, a batch of asset variations) and test whether ultra mode handles the decomposition well enough to drop your scaffolding.
- Lean on caching. With cache reads 90% off, move your stable system prompt and shared context into cached blocks so the repeated parts of every call cost almost nothing.
What to do next
GPT-5.6 is preview-only for now, so most teams will not get hands-on access immediately. The useful move today is to design for the tiered structure: build your app so the model and effort level for each step is a config value, not a hardcoded string. When Sol, Terra, and Luna reach general availability in the coming weeks, you flip the config instead of rewriting your integration, and you can A/B the tiers against your current model on cost and quality.
Frequently asked questions
What is GPT-5.6?
GPT-5.6 is OpenAI's next-generation model family announced on June 26, 2026. It ships as three models, Sol, Terra, and Luna, in a limited preview inside Codex and the API, with broader availability planned for the coming weeks.
What is the difference between Sol, Terra, and Luna?
They are capability tiers. Sol is the flagship for the hardest reasoning, Terra is the balanced everyday model, and Luna is the fast, affordable option. OpenAI describes them as durable tiers that advance on their own cadence, separate from the generation number.
What is GPT-5.6 ultra mode?
Ultra mode uses subagents to handle complex work beyond what a single agent can do, orchestrating multiple internal helpers on long, multi-part tasks instead of answering in a single pass.
How much does GPT-5.6 cost?
Per 1M tokens: Sol is $5 input / $30 output, Terra is $2.50 / $15, and Luna is $1 / $6. Cache writes are 1.25x the input rate and cache reads are discounted 90%.
When can I use GPT-5.6?
It is in limited preview for trusted partners and organizations now. OpenAI plans to make Sol, Terra, and Luna generally available in the coming weeks, with a staggered rollout tied to U.S. government coordination.
How is GPT-5.6 better than GPT-5.5?
OpenAI highlights advances in coding, biology, and cybersecurity, plus the new max reasoning effort and ultra orchestration mode. The company did not publish head-to-head benchmarks against GPT-5.5 in the announcement.
Sources: OpenAI: Previewing GPT-5.6 Sol, 9to5Mac, Thurrott, Tech Times, OpenAI: Introducing GPT-5.5.
