OpenAI introduces the flagship GPT-5.6 lineup: Sol, Terra, and Luna — a new perspective on AI models
OpenAI has taken a significant step in the development of its technologies by announcing the GPT-5.6 family of models, which includes three versions: the flagship Sol, the versatile Terra, and the budget-friendly Luna. Currently, access to them via the API and Codex has been granted to a limited circle of trusted partners. The company emphasizes that this launch was coordinated with U.S. authorities, who requested a limited preview format for safety assessment.
Key Features and Positioning
Sol is positioned as the most powerful model in OpenAI's history. It is equipped with two innovative modes: max for deep analysis of complex tasks requiring extended reasoning, and ultra, which uses sub-agents to accelerate the execution of multi-step processes. Terra, according to the developers, demonstrates performance comparable to GPT-5.5 but at half the cost. Luna is the fastest and most economical option, designed for everyday tasks.
Benchmarks and Performance
The test results are impressive. Sol set a new record in Terminal-Bench 2.1 for command-line tasks. On GeneBench v1, the model outperformed GPT-5.5 while using fewer tokens. In cybersecurity tasks on ExploitBench, Sol proved competitive with Mythos Preview, using roughly one-third of the output tokens. All three models showed improved results in cyber tasks on ExploitGym as reasoning depth increased.
Safety and Control
OpenAI states that GPT-5.6 Sol does not cross the Cyber Critical threshold within its Preparedness Framework. Tests showed that the model can find bugs and exploit primitives in browsers (Chromium, Firefox) but could not autonomously create a fully functional exploit. Over 700,000 GPU-hours (in A100 equivalent) were used for automated red-teaming to verify protections. Multi-layered measures have been introduced during the preview phase, including real-time monitoring and account-level restrictions.
Pricing for the models is set as follows: Sol — $5 per 1 million input tokens and $30 per output tokens; Terra — $2.50 and $15; Luna — $1 and $6. In July, OpenAI plans to launch Sol on the Cerebras platform, which will provide speeds of up to 750 tokens per second, but access at this stage will also be limited.
Expert Insight: The launch of GPT-5.6 is not just an update but a strategic rethinking of the market. The division into three clearly segmented models (flagship, workhorse, and budget option) is a direct response to growing competition from DeepSeek and other players. The emphasis on cybersecurity and collaboration with authorities deserves special attention. This signals to the market that OpenAI is ready for stricter regulation but aims to maintain leadership in innovation by offering tools for any task and budget.