OpenAI Announces Unlimited ChatGPT Text Chats and Updates GPT-5.6 Sol
OpenAI said on August 6 that ChatGPT Free and Go users will receive unlimited text chats powered by GPT-5.6 Luna starting the following week, while uploads, images, and other tools keep separate limits. Plus and Pro users now have an updated GPT-5.6 Sol and a reasoning slider; OpenAI says an internal three-domain evaluation found factual-error-bearing answers were 68% less common than with GPT-5.5 Instant.
OpenAI said on August 6 that ChatGPT Free and Go users will receive unlimited text conversations powered by GPT-5.6 Luna starting the following week. OpenAI's announcement and Axios's exact-event report say uploads, images, and other tools will continue to have separate limits.
Free and Go users will also receive a Think button that gives GPT-5.6 Luna more time for harder questions. OpenAI's current help documentation confirms that Think uses Luna rather than GPT-5.6 Sol, which remains limited to eligible paid plans in standard ChatGPT conversations. The announcement describes unlimited text access and the Think button as a scheduled rollout subject to abuse guardrails.
GPT-5.6 Sol changes for paid plans
For Plus and Pro subscribers, OpenAI updated GPT-5.6 Sol in regular ChatGPT conversations on August 6. OpenAI and Axios describe the revised model as designed to produce more direct answers, use tighter formatting, and avoid detail that does not help. OpenAI also says it should make fewer mistakes when answers depend on dates, numbers, sources, rules, or assumptions.
In an internal evaluation spanning financial, medical, and legal prompts, OpenAI says answers containing at least one factual error were 68% less common with the revised GPT-5.6 Sol than with GPT-5.5 Instant. This is a vendor-reported aggregate result, not an independently published benchmark, and it does not establish performance on a team's proprietary data or task distribution.
Plus and Pro users can also use a slider on ChatGPT web, mobile, and desktop to control how much reasoning an answer receives. OpenAI's help documentation lists Medium and High Sol reasoning for Plus, with Extra High and Pro options available on higher tiers. The August 6 update applies to standard ChatGPT conversations; OpenAI says the Sol versions used in Work and Codex are not changing as part of this release.
What teams should validate
Unlimited text access lowers the friction of testing prompt flows and longer conversations on consumer accounts, but upload, image, and tool limits remain separate constraints for multimodal work. The split also matters when reproducing behavior: Free and Go text chats use Luna, while the paid-plan reliability claim concerns the updated Sol model.
For research, policy, and code-adjacent workflows, teams should test date-sensitive retrieval, numerical reasoning, source attribution, and failure handling against representative internal examples. The 68% figure is useful evidence about OpenAI's test set, but it should be treated as a starting hypothesis for evaluation rather than a deployment guarantee.
Key Points
- 1Free and Go users are scheduled to receive unlimited GPT-5.6 Luna text chats the week after OpenAI's August 6 announcement, while uploads, images, and other tools retain separate limits.
- 2Plus and Pro users received an updated GPT-5.6 Sol and reasoning slider; OpenAI reports 68% fewer error-bearing answers in an internal three-domain evaluation.
- 3Teams should evaluate the updated model on representative tasks because a vendor aggregate does not establish reliability on proprietary data or workflows.
Scoring Rationale
The update changes text-chat access for ChatGPT's large Free and Go user base and changes reasoning controls for paid users. The reported GPT-5.6 Sol reliability improvement matters to factual and research workflows, but it remains a vendor internal evaluation.
Sources
Primary source and supporting public references used for this report.
Practice interview problems based on real data
1,625 SQL & Python problems across 15 industry datasets — the exact type of data you work with.
Try 250 free problems
