OpenAI Launches GPT-6 Sol and Luna, Cuts API Prices by 50%

OpenAI released GPT-6 Sol and Luna, cutting API prices in half and claiming fewer factual mistakes than its earlier models.

maisiekooc
Maisie Morrison

AgentLocker Editor

AI News
OpenAI Launches GPT-6 Sol and Luna, Cuts API Prices by 50%

OpenAI has released two new AI models, GPT-6 Sol and GPT-6 Luna. The launch came on September 22, 2026, and expands the GPT-6 family the company started earlier this month.

In early September, OpenAI launched GPT-6 Astra. The company calls Astra its most intelligent model and says it remains its best option across the board.

Sol and Luna are smaller and cheaper. OpenAI says it trained them with methods similar to those used for Astra.

Prices Cut in Half

The biggest change is price. OpenAI is cutting API prices for both models by 50% compared with the GPT-5.6 promotional pricing.

GPT-6 Sol now costs $2 per million input tokens and $10 per million output tokens. That is down from $4 and $20.

GPT-6 Luna costs $0.10 per million input tokens and $0.50 per million output tokens. The old prices were $0.20 and $1.20.

OpenAI said better caching and inference let it serve the models at lower cost. The company said it is passing those savings on to customers.

Sol is built for complex work such as coding. Luna is meant for high-volume tasks with a clear goal, like summarizing documents or answering quick questions.

Claims of Fewer Mistakes

OpenAI says GPT-6 Sol makes about half as many factual mistakes as the model before it. The company based this on an internal test using real conversations where users had flagged errors.

The company also compared its models with those from rival Anthropic. On a business workflow test called AutomationBench, OpenAI said Sol scored 33.2% at a cost of $0.27 per task.

By comparison, OpenAI said Anthropic's Claude Opus 5 scored 26.9% on the same test. It said Opus 5 cost more than 11 times as much per task.

On DeepSWE, a coding test, OpenAI said Sol scored 68.8%. That was within 1.1 percentage points of Anthropic's Claude Fable 5, at about 80% lower cost per task.

OpenAI said Luna scored 66.6% on the same coding test. It said that result was comparable to Claude Opus 5 and Fable 5 at medium effort, at a much lower cost.

All of these results come from OpenAI's own announcement. The company also said both models now write with more clarity, less jargon and slightly shorter answers.

Anthropic released a new version of its model, Claude Opus 5.5, about 90 minutes before OpenAI's announcement, according to TechCrunch.

The new Sol and Luna models are available now in ChatGPT Work and Codex for most paid accounts, and in the API. Luna is also coming to the desktop app and to Free and Go users, with OpenAI rolling the models out to the ChatGPT app and website gradually throughout the day.

From our research desk
AI Jobs Automation Index
Which jobs are AI tools targeting most? We mapped 3,400+ AI tools to real job functions — with BLS employment & salary data.
Explore the index
maisiekooc

Written by

Maisie is a news writer at Agent Locker, covering the latest developments in artificial intelligence, emerging technology and the companies shaping the future.

Discover AI Agents