A Three-Tier Revolution
On July 9, 2026, OpenAI released its most anticipated model family yet: the GPT-5.6 series, comprising three distinct tiers — Sol, Terra, and Luna — named after the Sun, Earth, and Moon. The release marks a strategic shift from a single monolithic model to a tiered architecture where each variant is optimized for different workloads, price points, and capability requirements.
Sol is the flagship — OpenAI's most powerful model ever, designed for complex coding, scientific research, biology workflows, and cybersecurity. Terra offers competitive performance to GPT-5.5 at half the cost, positioning it as the everyday workhorse. Luna is the budget champion, delivering strong capability at the lowest price point yet.
Benchmark-Leading Capabilities
GPT-5.6 Sol sets new state-of-the-art results on Terminal-Bench 2.1, a benchmark testing real-world command-line workflows requiring planning, iteration, and multi-tool coordination. It also achieves top scores on GeneBench v1 for long-horizon genomics analysis, and competitive results on ExploitBench for cybersecurity vulnerability research — using roughly one-third the output tokens of comparable models.
A new "max" reasoning effort mode gives Sol additional thinking time for the hardest problems, while an "ultra" mode deploys sub-agents that collaborate on complex tasks, pushing beyond what a single model instance can achieve.
Safety First: The Most Robust Stack Ever
Perhaps the most significant aspect of the launch is OpenAI's unprecedented emphasis on safety. The company dedicated over 700,000 A100-equivalent GPU hours to automated red-teaming, searching for universal jailbreaks that work across many prompts. This was complemented by extensive human expert red-teaming.
The safeguard stack includes: - Model-level training to refuse prohibited cyber and biology assistance - Real-time output classifiers that can pause generation mid-stream for review by a larger reasoning model - Account-level behavioral analysis spanning multiple conversations - Differentiated access controls for sensitive capabilities
The release followed a 13-day government preview period where OpenAI shared model plans and capabilities with U.S. authorities — a process OpenAI says should not become the long-term default but was necessary for broader availability.
Pricing and Availability
GPT-5.6 is priced per million tokens across three tiers: Sol at $5/$30 (input/output), Terra at $2.50/$15, and Luna at $1/$6. The models are initially available through API and Codex to select partners, with broader ChatGPT, Codex, and API access planned in the coming weeks. Notably, Sol will also be available on Cerebras hardware at up to 750 tokens per second.
The naming convention itself signals OpenAI's new philosophy: the generation number (5.6) identifies the model family, while Sol, Terra, and Luna represent durable capability tiers that can advance on their own cadence — meaning future updates won't necessarily require a full generation jump.
Industry Impact
The launch arrives amid intensifying competition in frontier AI. With Sol's demonstrated gains in coding, biology, and cybersecurity — and a pricing structure designed to capture both enterprise and developer markets — OpenAI is positioning the GPT-5.6 family as the most versatile AI platform available, while signaling that safety and government engagement will remain central to its strategy going forward.
