HasItGot
Tech

OpenAI Launches Budget-Friendly GPT-6 Sol and Luna Models

OpenAI has introduced GPT-6 Sol and Luna, bringing lower token pricing, faster responses, and stronger coding capabilities to developers.

HasItGot Newsroom
OpenAI Launches Budget-Friendly GPT-6 Sol and Luna Models

OpenAI has released GPT-6 Sol and GPT-6 Luna, adding two economical options to its lineup following the debut of GPT-6 Astra. The company applied training methods derived from Astra to these models, targeting better results across everyday office duties, coding tasks, factual precision, alignment, and automated computer operation.

Each model fills a distinct niche in the catalog. GPT-6 Sol is tailored for general professional tasks and carries deeper reasoning power than its sibling, while remaining less expensive than Astra. Meanwhile, GPT-6 Luna is built for high-throughput environments where speed is critical, serving as the most affordable option in the new family.

Reduced Pricing and Caching Improvements

Both models arrive with steep price cuts relative to the previous 5.6 generation. For GPT-6 Sol, API rates are halved, coming in at $2 per million input tokens and $10 per million output tokens. GPT-6 Luna is priced at $0.10 per million inputs alongside $0.50 per million outputs, dropping down from previous figures of $0.20 and $1.20.

OpenAI also upgraded prompt caching infrastructure to improve overall efficiency. Autonomous agents can retain broader context and respond more swiftly, with developers receiving a 90 percent discount whenever cached input tokens are read.

Benchmarking and Output Adjustments

Internal evaluations show that GPT-6 Sol cuts errors roughly in half compared to GPT-5.6 Sol. The model demonstrates clear progress in software development and direct computer utilization, equaling or exceeding Claude Fable 5.1 across select testing suites.

As reported by MacRumors, the competitive landscape shifted on the same day. While OpenAI compared Sol against older Opus 5 metrics, rival firm Anthropic rolled out Opus 5.5, which overtakes Astra on specific knowledge and programming tests. GPT-6 Sol costs half as much per token as Opus 5.5, though Anthropic maintains that its model requires fewer tokens to finish identical assignments.

In addition to raw technical gains, OpenAI modified how Sol and Luna communicate. Adopting the conversational adjustments introduced with Astra, both models generate slightly shorter answers that strip out unnecessary technical jargon and low-value context without sacrificing meaning. Safety assessments also noted a decrease in misleading statements regarding coding progress.

Rollout and Platform Availability

Subscribers on Plus, Pro, Business, Enterprise, and Edu tiers can begin using GPT-6 Sol and GPT-6 Luna immediately inside ChatGPT Work and Codex. Users on Free and Go accounts can access GPT-6 Luna exclusively within the desktop client. Neither model has been integrated into standard Chat mode yet.

More tech

All tech