GPT-5.6 is the right next watch, but the clean headline is narrower than the hype cycle will want: OpenAI has started a limited preview of the GPT-5.6 family, and the family currently means Sol, Terra, and Luna.
That is already enough to track. It is not enough to call it a broad ChatGPT release.
Current status
OpenAI's help article says GPT-5.6 is available during the preview through the OpenAI API and Codex for a limited group of trusted partners and organizations. The same article says GPT-5.6 is not available in ChatGPT during the preview, and that OpenAI has not announced a general-availability date.
That makes Hugin's current status label simple:
- Confirmed: GPT-5.6 exists as an official limited preview.
- Confirmed: Sol, Terra, and Luna are the named models in the family.
- Confirmed: Preview access can include API, Codex, or both, depending on the organization approval.
- Not confirmed: a public self-service signup, public waitlist, or broad ChatGPT rollout.
- Not confirmed: a dated general-availability launch.
The three versions
The short version:
- GPT-5.6 Sol is the flagship model. OpenAI frames Sol as the strongest GPT-5.6 model, with focus on coding, professional work, science, and cybersecurity.
- GPT-5.6 Terra is the middle lane. OpenAI describes it as the lower-cost strong option in the family.
- GPT-5.6 Luna is the fast lane. OpenAI describes Luna as the fastest and most cost-efficient GPT-5.6 model.
The pricing receipt is already public in the help article: gpt-5.6-sol is
listed at $5 per 1M input tokens and $30 per 1M output tokens, gpt-5.6-terra
at $2.50 input and $15 output, and gpt-5.6-luna at $1 input and $6 output.
The same page says GPT-5.6 introduces explicit cache breakpoints and a
30-minute minimum cache life, with cache writes billed at 1.25x uncached input
and cache reads keeping the 90% cached-input discount.
Why this is the next tracker
Fable 5 was a restoration story. Sonnet 5 was a model-picker story. GPT-5.6 is different because it is a staged release story with multiple product lanes and a safety record attached from day one.
OpenAI's launch post says GPT-5.6 introduces max reasoning effort and an
ultra mode that uses subagents for complex work. That makes the public
question less "is the model bigger?" and more "where can people use it, what
does each lane cost, how much workflow evidence is visible, and what safeguards
shape the work?"
The system card is the other anchor. It says GPT-5.6 is a three-model family and describes the launch safeguards as OpenAI's most robust yet. That matters because the broad-access headline should wait until the access language, docs, status pages, pricing, and safety record all line up.
What Hugin should watch next
The next real receipts are practical:
- A dated OpenAI post or help-center update that changes "limited preview" into broader availability.
- API documentation showing the model IDs as generally usable outside selected organizations.
- Codex release notes showing which workspaces or plans receive Sol, Terra, or Luna.
- ChatGPT release notes showing whether GPT-5.6 enters the consumer model picker.
- Pricing-page updates that match or revise the help-center prices.
- Status-history notes if availability, safeguards, or preview access create visible service interruptions.
Until those receipts appear, the honest headline is: GPT-5.6 is real, public details exist, and broad access is still pending.