OpenAI has officially launched GPT-6 Sol and GPT-6 Luna, two new API-only models within its GPT-6 model family. Positioned below the flagship GPT-6 Astra released earlier this month, Sol targets complex professional and coding tasks, while Luna is built for high-volume, everyday workloads. Both models are immediately deployable via the OpenAI API.
The release centers heavily on cost efficiency and performance improvements. Supported by enhanced inference and caching architectures, OpenAI has reduced API pricing for Sol and Luna by at least 50% compared to GPT-5.6 promotional rates, with Luna’s output pricing dropping 58% to $0.50. On benchmarks such as AutomationBench and DeepSWE v1.1, Sol and Luna match or exceed competing models like Claude Opus 5 at significantly lower task costs.
Developer optimizations include an upgraded prompt caching system offering up to 90% discounts on cached input reads within a 30-minute window. Early integration data from GitHub Copilot indicates the new caching structure cuts fresh prompt token processing by more than 50%, enabling faster response times.
Why it matters
Developers can reduce production API costs by 50% or more while maintaining high-level reasoning and coding performance.
Improved prompt caching lowers latency and overhead for complex agentic workflows with repeating instructions.
OpenAI is aggressively targeting enterprise mid-tier workloads to undercut competitor pricing across agent and coding tasks.
Source: marktechpost.com



