OpenAI has released its GPT-5.6 model family, introducing two distinct models: Luna and Sol. Luna is priced 80% lower than comparable models, while Sol leads on benchmarks for reasoning and cost efficiency.
Two Models, Two Strengths
The GPT-5.6 family splits into Luna and Sol, each targeting different priorities. Luna's pricing undercuts comparable models by 80%, making it a budget-friendly option for developers and businesses. Sol, on the other hand, tops internal benchmarks in reasoning and cost efficiency, positioning it as a high-performance choice.
Pricing Strategy
Luna's 80% price reduction is a sharp departure from typical AI model pricing. The lower cost could open access to advanced language capabilities for smaller teams or projects with tight budgets. OpenAI did not disclose the exact price per token or the specific models Luna is compared against.
Benchmark Performance
Sol leads on benchmarks for reasoning and cost efficiency, according to OpenAI. The company has not released the full benchmark results or the methodology used. The model's performance suggests it may be suited for complex tasks where accuracy and resource use matter most.
Both models are now part of OpenAI's lineup, though the company has not detailed availability or deployment options. Developers will have to weigh Luna's cost advantage against Sol's benchmark leadership.
