Last updated: July 7, 2026

ChatGPT 5.6 is the phrase many people will search for, but the official product family is GPT-5.6. OpenAI previewed three GPT-5.6 models: Sol, Terra, and Luna. Sol is the flagship model, Terra is the balanced lower-cost option, and Luna is the fastest and most cost-efficient option.

Quick answer: GPT-5.6 is currently a limited preview, not a normal ChatGPT rollout. OpenAI’s Help Center says Sol, Terra, and Luna are available during the preview through the OpenAI API and Codex for a limited group of trusted partners and organizations. It also says GPT-5.6 is not available in ChatGPT during the preview and that OpenAI has not announced a general-availability date.

Source check Sources checked

Verified July 7, 2026

Important: this page covers the preview period. Model access, ChatGPT availability, limits, and public rollout details may change after OpenAI publishes a general-availability update.

GPT-5.6 model family overview

Try it in practice Make this section actionable Practice the workflow instead of only comparing tools.
Model Model ID Role Input Output Best fit
GPT-5.6 Sol gpt-5.6-sol Flagship $5 / 1M $30 / 1M Hard reasoning, agentic coding, science, cybersecurity, complex workflows
GPT-5.6 Terra gpt-5.6-terra Balanced $2.50 / 1M $15 / 1M Everyday professional work, coding help, analysis, routing default
GPT-5.6 Luna gpt-5.6-luna Fast/low-cost $1 / 1M $6 / 1M High-volume tasks, drafts, extraction, classification, cheap first pass

For deeper pages, use:

ChatGPT 5.6 release date and access

Try it in practice Make this section actionable Practice the workflow instead of only comparing tools.

The official GPT-5.6 preview announcement was published on June 26, 2026. That is the date OpenAI previewed the model family publicly.

But release date does not mean broad access. During the preview:

  • approved organizations may get access through the OpenAI API, Codex, or both;
  • approval for API access does not automatically mean Codex access;
  • approval for Codex does not automatically mean API access;
  • the preview is not a public self-service signup;
  • there is no public application or waitlist;
  • GPT-5.6 is not available in ChatGPT during the preview.

This matters because screenshots or social posts claiming “ChatGPT 5.6 access” can be misleading. The official preview access path is API/Codex for selected organizations, not broad ChatGPT consumer access.

Sol vs Terra vs Luna: which one should you test?

Try it in practice Make this section actionable Practice the workflow instead of only comparing tools.

Think of GPT-5.6 as a routing family.

If your task is… Start with Why
A difficult autonomous coding workflow Sol Higher capability and deeper reasoning matter more than cost.
A normal software planning or analysis task Terra Better cost balance than Sol.
High-volume summaries or classification Luna Cheapest and fastest GPT-5.6 tier.
Security-sensitive review Sol or Terra, with strict policy controls Safety and evaluation matter more than speed.
Drafting, extraction, routing, tagging Luna Often enough for simple structured work.
Executive research synthesis Terra first, Sol for final hard cases Start balanced, escalate only when needed.
A consumer ChatGPT workflow None during preview OpenAI says GPT-5.6 is not in ChatGPT during the preview.

A good production setup probably will not use Sol for everything. It will route easy tasks to Luna, normal professional tasks to Terra, and difficult high-value tasks to Sol.

GPT-5.6 pricing and cache math

Try it in practice Make this section actionable Practice the workflow instead of only comparing tools.

OpenAI lists GPT-5.6 pricing per 1M tokens:

Model Input Output Cache write Cached input read
GPT-5.6 Sol $5.00 $30.00 $6.25 $0.50
GPT-5.6 Terra $2.50 $15.00 $3.125 $0.25
GPT-5.6 Luna $1.00 $6.00 $1.25 $0.10

The cache-write and cached-input figures come from OpenAI’s rule for GPT-5.6 and later models:

  • cache writes are billed at 1.25x the uncached input rate;
  • cached-input reads receive a 90% discount;
  • GPT-5.6 introduces more predictable prompt caching with explicit cache breakpoints and a minimum cache life of 30 minutes.

For real applications, compare cost per successful task, not only price per token. If Sol solves a task in one attempt that Luna fails five times, Sol may be cheaper overall. If Luna handles 90% of traffic reliably, routing everything to Sol wastes budget.

What is new in GPT-5.6?

Try it in practice Make this section actionable Practice the workflow instead of only comparing tools.

OpenAI’s launch post frames GPT-5.6 around stronger capability and a more robust safety stack. The main themes are:

  • agentic coding and command-line workflows;
  • professional knowledge work;
  • computer use;
  • scientific research;
  • cybersecurity and defensive security workflows;
  • max reasoning effort for Sol;
  • ultra mode that can use subagents for complex work;
  • stronger safeguards for higher-risk cyber and biology requests.

The detailed GPT-5.6 Sol page focuses on the flagship model. This hub is for choosing among all three tiers.

GPT-5.6 preview caveats

Try it in practice Make this section actionable Practice the workflow instead of only comparing tools.

Before you plan a migration, keep these limits in mind:

  1. No broad ChatGPT access during preview. OpenAI’s Help Center says ChatGPT is not included.
  2. No public self-service preview signup. Organizations need selected access.
  3. Access may be scoped. API and Codex approvals are separate.
  4. Safety checks may add latency. Higher-risk biology or cyber prompts can be blocked or slowed by extra checks.
  5. General availability is not dated. OpenAI says it plans to expand availability, but has not announced a GA date.
  6. The API models page still points most developers to GPT-5.5. It notes GPT-5.6 preview availability for select trusted partners.

Evaluation checklist

Try it in practice Make this section actionable Practice the workflow instead of only comparing tools.
Evaluation template GPT-5.6 routing scorecard
Question Why it matters Likely model
Does failure cost a lot? Hard tasks may justify a flagship tier. Sol
Is the task routine but still professional? Balanced quality/cost matters. Terra
Is the task high-volume and easy to verify? Cheap routing can save large budgets. Luna
Does the prompt include security or biology content? Extra safeguards and review may apply. Sol/Terra with controls
Is the user waiting interactively? Latency matters. Luna or Terra
Does the task need subagents or ultra mode? OpenAI describes ultra mode around Sol. Sol

Bottom line

Try it in practice Make this section actionable Practice the workflow instead of only comparing tools.

GPT-5.6 is a three-tier family, not one model:

  • Sol: highest capability, highest cost;
  • Terra: balanced everyday model;
  • Luna: fastest and most cost-efficient model.

During the current preview, GPT-5.6 is mainly a tool for selected API and Codex partners, not a consumer ChatGPT release. The right move is to prepare routing tests now, then migrate only after your organization has confirmed access, pricing, latency, safety behavior, and workload-specific quality.

FAQ

Try it in practice Make this section actionable Practice the workflow instead of only comparing tools.
What is ChatGPT 5.6?
ChatGPT 5.6 is the common search name for OpenAI’s GPT-5.6 model family. The official model family includes GPT-5.6 Sol, GPT-5.6 Terra, and GPT-5.6 Luna.
Is GPT-5.6 available in ChatGPT?
No, not during the preview according to OpenAI’s Help Center. During the preview, approved participants may access GPT-5.6 through the OpenAI API, Codex, or both.
What are the GPT-5.6 model IDs?
OpenAI lists gpt-5.6-sol, gpt-5.6-terra, and gpt-5.6-luna in its GPT-5.6 preview pricing table.
How much does GPT-5.6 cost?
OpenAI lists GPT-5.6 Sol at $5 input and $30 output per 1M tokens, Terra at $2.50 input and $15 output, and Luna at $1 input and $6 output.
Which GPT-5.6 model should I use?
Use Sol for high-value hard tasks, Terra for balanced everyday professional work, and Luna for fast high-volume tasks that are easy to verify.