OpenAI Launches Unlimited Free GPT-5.6 Luna as Default ChatGPT Model
OpenAI announced Thursday that its GPT-5.6 Luna model will become the default engine for non-paying ChatGPT users, lifting daily message caps for standard text conversations. The change applies to Free and Go tier accounts, removing a restriction that previously halted lengthy user sessions. The platform currently serves more than one billion weekly active users.
According to an official company announcement, the model replacement is rolling out to Free and Go accounts this week.
Unlimited text conversations and a new dedicated "Think" feature will arrive during the week of August 10.
The company confirmed that standard quotas will remain in place for image creation, voice interactions, file processing, and visual analysis.
Model Capabilities and Benchmark Performance
The new default, GPT-5.6 Luna, replaces GPT-5.5 Instant across entry tiers.
OpenAI internal benchmarks indicate that Luna reduces factual errors by 62% when compared to its predecessor across legal, medical, and financial testing prompts.
Users will also gain access to a "Think" button within the chat interface starting next week.
 |
| Credit: OpenAI/ChatGPT |
The feature instructs GPT-5.6 Luna to spend additional computing time working through multi-step logic before returning an answer.
On API documentation pages, OpenAI describes Luna as a lightweight model designed for high-volume, latency-sensitive workloads with a 1.05 million token context window.
Subscriber Upgrades and System Guardrails
Paid ChatGPT Plus and Pro subscribers are receiving an updated version of the GPT-5.6 Sol model. OpenAI tuned this release for directness, shorter answers, and cleaner formatting.
Corporate documentation shows that the updated Sol produced 68% fewer factual errors than GPT-5.5 Instant during internal evaluations.
Paid tiers also gain customizable reasoning controls to adjust processing depth on individual queries.
In an official update published to its help center, OpenAI stated:
"Go users can send up to 10 messages every 5 hours after enabling Thinking."
The company added that general usage caps continue to apply based on overall platform load and abuse prevention systems.
Under internal safety evaluations published by the company, OpenAI classified both GPT-5.6 Sol and GPT-5.6 Luna as high-capability systems in cybersecurity and biological testing domains.
System safeguards remain active across all user tiers to manage automated requests and prevent platform exploitation.
