OpenAI sets July 9th public launch for GPT-5.6 Sol, Terra and Luna
The rollout follows a government-coordinated limited preview that began June 26th and kept the models out of ChatGPT.
By Ryan Merket · Published
Why it matters
OpenAI is moving GPT-5.6 from a government-coordinated preview to public launch in under two weeks, testing how fast frontier models can ship when cyber-risk controls are part of the release plan.
OpenAI said GPT-5.6 Sol, Terra and Luna will launch publicly on Thursday, July 9th, moving its newest model family beyond the restricted preview it began less than two weeks ago.
The company posted the timing late on July 7th U.S. time, writing that it is also "expanding preview access globally now." OpenAI did not specify in the post whether the July 9th public launch means full access across ChatGPT, Codex and the API, or a staged release across those surfaces. Its existing June 26th launch announcement said the GPT-5.6 models would initially be available through the API and Codex to selected trusted partners, with broader availability for ChatGPT, Codex and API users planned soon.
The short gap between preview and public launch matters because OpenAI framed the first GPT-5.6 rollout as a policy-mediated release. In the June 26th post, OpenAI said it had previewed the models' capabilities and release plans to the U.S. government before launch, and that it was starting with a small group of trusted partners at the government's request. OpenAI also said it did not believe that kind of access process should become the long-term default.
GPT-5.6 is OpenAI's first model family to use the Sol, Terra and Luna naming scheme. OpenAI describes Sol as the flagship model, Terra as a balanced model for everyday work, and Luna as the fast, lower-cost option. The company says the number identifies the model generation, while the names identify capability tiers that can advance on separate cadences.
The pricing gives developers a clearer view of how OpenAI is segmenting the family. In the June 26th announcement and an accompanying Help Center article, OpenAI listed GPT-5.6 Sol at $5 per 1 million input tokens and $30 per 1 million output tokens. Terra is priced at $2.50 input and $15 output. Luna is priced at $1 input and $6 output.
That structure makes Terra the cost argument in the family. OpenAI says Terra has performance competitive with GPT-5.5 while being 2x cheaper, while Luna brings what OpenAI calls strong capability at its lowest cost. Those are company benchmarks, and OpenAI said it would publish an expanded suite of evaluation results when the models become broadly available.
Sol is the model OpenAI is using to make the capability claim. The company says GPT-5.6 Sol is its strongest model yet, with gains in coding, biology workflows and cybersecurity. OpenAI said Sol set a new state of the art on Terminal-Bench 2.1, a command-line workflow benchmark, and showed stronger results than GPT-5.5 on GeneBench v1 while using fewer tokens. On cyber tasks, OpenAI said Sol was competitive with Anthropic's Mythos Preview on ExploitBench while using roughly one-third of the output tokens.
The cyber emphasis also explains why OpenAI wrapped the launch in safeguards language. The company said GPT-5.6 Sol, Terra and Luna were developed with its most robust safeguards to date, including model-level refusals, real-time checks during generation, account-level review signals and differentiated access. OpenAI said the models may block or slow some legitimate requests during the preview, especially in dual-use biology and cybersecurity work, while the company tunes the safeguards.
OpenAI also said GPT-5.6 Sol does not cross the "Cyber Critical" threshold under its Preparedness Framework. The company's wording was careful: in Chromium and Firefox evaluations, Sol identified bugs and exploitation primitives, but did not autonomously produce a functional full-chain exploit under the tested conditions. OpenAI added that benchmark thresholds cannot capture every way a model may be combined with other tools.
The July 9th launch will test whether OpenAI can make that release posture work at public scale. During the limited preview, the Help Center said individual consumers were not eligible, paid ChatGPT plans did not provide access, there was no public application or waitlist, and approved organizations needed an OpenAI account representative. The public launch post changes the timeline, but not yet the implementation details.
The most concrete developer changes beyond model access are in caching and deployment. OpenAI said GPT-5.6 introduces explicit cache breakpoints and a 30-minute minimum cache life. Cache writes are billed at 1.25x the uncached input rate, while cache reads continue to receive a 90% cached-input discount. OpenAI also said GPT-5.6 Sol would launch on Cerebras in July at up to 750 tokens per second, initially for select customers as capacity expands.
OpenAI's public message on July 8th is therefore narrower than a full product brief, but the timing is clear. The restricted preview began on June 26th. The company is widening preview access globally on July 8th. The public launch is scheduled for Thursday, July 9th. The open questions are product-surface questions: whether ChatGPT users get access immediately, whether API access becomes self-service, and whether Sol, Terra and Luna all arrive with the same limits.