OpenAI released GPT-6 Sol and Luna on Tuesday at half the API price of the GPT-5.6 versions, and says the Sol model makes half as many mistakes as its predecessor on an internal factuality test

A frontier lab dropping its API prices by half two hours after a rival does the same is the industry saying that model choice is now a price decision. TechCrunch reports OpenAI released GPT-6 Sol and Luna on 22 September 2026, priced at half the API cost of the GPT-5.6 versions of the same names, extending the GPT-6 family that started with the flagship Astra earlier this month. The company attributes the price cut to caching and inference improvements.

What the vendor says the models do

TechCrunch quotes OpenAI's announcement that "GPT-6 Astra introduced a new generation of intelligence; these models extend its benefits by making that intelligence more efficient and accessible." The company also says, in its own announcement quoted by TechCrunch, "On our internal factuality evaluation, which is based on de-identified real-world conversations where users flagged mistakes by our models, GPT-6 Sol makes about half as many mistakes as its predecessor, reaching Astra-level reliability at much lower cost." That is OpenAI evaluating itself, and worth an independent check before it drives a purchasing decision.

OpenAI positions the two models against Anthropic. TechCrunch reports the launch repeatedly claims GPT-6 Sol and Luna handle tasks better than Anthropic's top models, including Fable and Opus. A vendor benchmarking itself against a competitor is marketing rather than measurement. TechCrunch notes Anthropic released Opus 5.5 90 minutes before OpenAI's release, priced at $4 per million input tokens and $20 per million output tokens.

Where to get them, and what tier gets what

GitHub confirmed the same models are available in Copilot the same day. The GitHub Changelog describes Sol as "a balanced model for interactive and agentic coding" and Luna as "a lightweight, cost-efficient model for smaller, faster tasks and the lowest-cost option in the GPT-6 family." Sol is available to Copilot Pro+, Max, Business and Enterprise plans. Luna also reaches Pro plans and is enabled by default unless an administrator turns it off. Both are billed under usage-based billing, which the GitHub docs explain in its Copilot pricing reference.

Directly with OpenAI, TechCrunch reports the models are available in ChatGPT Work and Codex for most paid accounts, and in the ChatGPT API, with Luna also reaching the desktop app and Free and Go tiers. Rollout to the ChatGPT web and app was expected to happen through the day.

The pattern that is now real

Two frontier labs cut API prices for the same class of model on the same afternoon, and one of them made a factuality claim large enough to move a purchasing decision if it holds up. That is a real pattern: agent-shaped workloads are now competitive on cost per task, and the labs are competing on the token count a workflow actually burns as well as on the per-token price. A team already running eval harnesses on either family has a low-cost afternoon in front of them.

The factuality number is the one to check first, since factuality was OpenAI's largest quantitative claim and the one most likely to change what a workflow trusts. Any team relying on Sol for questions where a hallucination has a cost should run the same de-identified conversation samples through the older and newer models and count the flagged answers themselves.

Source

Rebecca Bellan, TechCrunch, OpenAI launches GPT-6 Sol and Luna, boasting lower cost and fewer mistakes, 22 September 2026. Vendor announcement: OpenAI, Introducing GPT-6 Sol and Luna, 22 September 2026. Availability in GitHub Copilot: GitHub Changelog, OpenAI's GPT-6 Sol and GPT-6 Luna now available, 22 September 2026.