OpenAI Unveils GPT-5.6 Sol, Terra, and Luna in Limited Preview

OpenAI announced a limited preview of its next-generation GPT-5.6 model family today, introducing a new three-tier lineup led by the flagship Sol model, which delivers major gains in agentic coding alongside enhanced reasoning capabilities.



The GPT-5.6 series introduces three distinct capability tiers. Sol serves as the flagship reasoning model, while Terra provides a balanced option with performance comparable to GPT-5.5 at half the cost. Luna rounds out the lineup as a fast, affordable model for low-latency tasks. GPT-5.6 also introduces a new naming system in which the version number identifies the model generation, while Sol, Terra, and Luna represent long-term capability tiers that can evolve independently. All three models are expected to become broadly available through the API, ChatGPT, and Codex in the coming weeks following the preview period.


To handle highly complex workloads, OpenAI is introducing a new max reasoning effort setting that gives Sol more time to reason through difficult problems. The update also brings a new ultra mode that pushes beyond the capabilities of a single agent by leveraging multiple subagents to accelerate complex work in parallel. OpenAI says Sol sets a new state-of-the-art score on Terminal-Bench 2.1 for command-line workflows and demonstrates strong improvements in long-horizon cybersecurity tasks.

Broader availability of the GPT-5.6 series will be delayed slightly as OpenAI begins with a limited preview for a small group of trusted partners at the request of the U.S. government. The phased rollout follows discussions about the models' cyber capabilities and comes just weeks after the U.S. government required Anthropic to suspend its Claude Fable 5 and Claude Mythos 5 models over cybersecurity and jailbreak concerns. OpenAI said it does not believe this type of government review should become the long-term default but is taking the short-term step as it works toward broader availability in the coming weeks.

To harden the system, OpenAI devoted more than 700,000 GPU hours to automated red-teaming aimed at finding universal jailbreaks. The company also deployed real-time classifiers that can pause generations if potential cyber or biological misuse is detected, allowing a larger reasoning model to review the conversation and its context before a response is delivered.

When the models become broadly available, Sol will cost $5 per million input tokens and $30 per million output tokens. Terra is priced at $2.50 for input and $15 for output, while Luna costs $1 and $6, respectively. OpenAI is also introducing more predictable prompt caching with explicit cache breakpoints and a minimum 30-minute cache life. Cache writes will be billed at 1.25 times the uncached input rate, while cached reads continue to receive a 90 percent discount.


Please download the iClarified app or follow iClarified on X, Facebook, YouTube, and RSS for more updates on availability.