OpenAI launches GPT‑6 Sol and Luna: frontier‑level AI at a bargain price

OpenAI has introduced two new large language models, GPT‑6 Sol and GPT‑6 Luna, positioned as cheaper, lighter variants of its flagship GPT‑6 Astra system. Announced on September 21, the models target organizations that want much of Astra’s power for coding and knowledge‑heavy tasks, but at a fraction of the cost.

According to tech outlets covering the release, Sol and Luna inherit core architecture from Astra but are optimized for efficiency: faster inference, lower compute requirements, and pricing tuned for everyday workloads. They’re described as “mini‑Astra” models, offering strong performance on software development, analysis, and general assistant use while staying budget‑friendly for startups and corporate IT teams.

The launch fits OpenAI’s broader strategy in 2026: moving beyond a single frontier model to a layered portfolio of systems matched to specific use cases and price points. The key questions now are how quickly customers migrate from older GPT‑4/5‑class models, and how OpenAI will differentiate Astra’s role as a cutting‑edge research‑grade system from Sol and Luna as workhorse models for large‑scale deployment.