Publication date: July 9, 2026
OpenAI has made its GPT-5.6 model family generally available, opening public access to GPT-5.6 Sol, Terra, and Luna after a limited preview that began on June 26, 2026. GPT-5.6 replaces OpenAI’s single-model approach with three durable capability tiers: Sol as the flagship, Terra as a balanced everyday model priced at roughly half of GPT-5.5, and Luna as the fastest and most affordable option. The release introduces a new max reasoning effort and an ultra mode that uses subagents for complex work.
What is GPT-5.6 Sol, Terra and Luna?
GPT-5.6 uses a new naming system in which the number identifies the model generation, while Sol, Terra, and Luna identify durable capability tiers that can advance on their own cadence. Sol is OpenAI’s most capable model, Terra targets everyday production work with GPT-5.5-class performance at lower cost, and Luna brings strong capability at OpenAI’s lowest price point.
The release adds two new controls. A max reasoning effort gives Sol the most time to reason deeply, and a new ultra mode goes beyond a single agent by leveraging subagents to accelerate complex, long-horizon work. OpenAI describes GPT-5.6 Sol as its strongest model yet for coding, science, and cybersecurity workflows.
GPT-5.6 benchmarks and technical specs
On Terminal-Bench 2.1, which tests command-line workflows requiring planning, iteration, and tool coordination, Sol Ultra scored 91.9% and Sol scored 88.8%, ahead of Claude Mythos 5 at 88.0% and Claude Opus 4.8 at 78.9%. Luna scored 82.5%, while Terra reached 84.3%, matching Claude Fable 5 and edging GPT-5.5 at 83.4%.
OpenAI also reported gains in biology and cybersecurity. On GeneBench v1, which evaluates long-horizon genomics and quantitative-biology analyses, GPT-5.6 Sol beat GPT-5.5 while using fewer tokens. On ExploitBench, Sol was competitive with Mythos Preview using only about one third of the output tokens. OpenAI states that GPT-5.6 Sol does not cross the Cyber Critical threshold under its Preparedness Framework: in tests on Chromium and Firefox it found bugs and exploitation primitives but did not autonomously produce a functional full-chain exploit. The context window was not officially confirmed at preview; GPT-5.5 shipped a 1 million token context window.
How does GPT-5.6 compare to Claude Opus 4.8 and GPT-5.5?
On Terminal-Bench 2.1, the GPT-5.6 tiers sit at or above the strongest Claude models: Sol Ultra at 91.9% and Sol at 88.8% lead Claude Mythos 5 at 88.0% and Opus 4.8 at 78.9%, while Terra at 84.3% matches Claude Fable 5. Against its own predecessor, Terra delivers GPT-5.5-class results at roughly half the cost, and Luna opens a cheaper tier below any prior OpenAI production model.
The efficiency story is central to OpenAI’s positioning. GPT-5.6 Sol reached its biology and cybersecurity results using fewer tokens than GPT-5.5 and than competing models on the same tasks, which lowers the effective cost of long-horizon agentic runs.
GPT-5.6 availability and pricing
GPT-5.6 Sol, Terra, and Luna are becoming generally available across ChatGPT, Codex, and the API following the preview period. During the preview the models were limited to a small group of trusted partners as part of OpenAI’s coordination with the U.S. government on a cyber Executive Order framework; the U.S. Department of Commerce subsequently approved a broader launch.
Pricing is per one million tokens across the three tiers: Sol is $5 input and $30 output, Terra is $2.50 input and $15 output, and Luna is $1 input and $6 output. GPT-5.6 introduces more predictable prompt caching with explicit cache breakpoints and a 30-minute minimum cache life; cache writes are billed at 1.25x the uncached input rate, while cache reads keep the 90% cached-input discount. OpenAI is also launching GPT-5.6 Sol on Cerebras at up to 750 tokens per second, initially for select customers.
Conclusion
With GPT-5.6 Sol, Terra, and Luna, OpenAI moves to a three-tier model family spanning intelligence, speed, and cost, adding a max reasoning effort and an ultra subagent mode.
Full details are in OpenAI’s announcement and the GPT-5.6 preview system card. Official announcement: https://openai.com/index/previewing-gpt-5-6-sol/
