OpenAI has expanded its GPT-6 family with GPT-6 Sol and GPT-6 Luna, two models designed to bring capabilities from the more powerful GPT-6 generation into lower-cost, higher-volume workloads.

OpenAI says both models were released on September 22, 2026, through its API, ChatGPT and Codex.

GPT-6 Sol vs GPT-6 Luna

The two models target different workloads.

GPT-6 Sol is designed for more demanding tasks such as coding, reasoning and computer use.

GPT-6 Luna focuses more on fast, high-volume tasks such as summarization, extraction and routine knowledge work.

OpenAI's published API pricing also shows a substantial difference between the two.

GPT-6 Sol is listed at $2 per million input tokens and $10 per million output tokens, while GPT-6 Luna is priced at $0.10 per million input tokens and $0.50 per million output tokens for the standard pricing tier described in the API changelog.

Why does this matter?

The important development is not simply another model launch.

It is model specialization.

Businesses do not need their most expensive model to handle every request.

A customer-support workflow might use a smaller model for routine questions and route complex cases to a stronger reasoning model.

The same architecture can work for document processing, research, coding and AI agents.

What should businesses consider?

When evaluating GPT-6 Sol and GPT-6 Luna, organizations should look beyond benchmark scores.

Consider:

Accuracy on your own data

Latency

Token costs

Tool-calling reliability

Context requirements

Output consistency

Security and governance

The GPT-6 Sol and Luna launch reflects a broader industry trend: AI infrastructure is moving toward intelligent model routing rather than using one model for everything.