OpenAI has released an update for GPT-5.6, upgrading the default model to GPT-5.6 Luna and introducing a more direct, concise conversation style (GPT-5.6 Sol). The new model significantly improves factual accuracy, reducing error rates by 68% compared to GPT-5.5, and adds a thinking depth slider and Think button, along with enhanced safety measures for minors.
SynthePulse Insight · AI deep reading
Version 1 · 1 source
OpenAI releases GPT-5.6 update, upgrading default model to Luna for free users, adding thinking depth slider and Think button, and strengthening minor protections. However, the news comes from third-party reposts, and specific features still need official page confirmation.
According to a repost by X user 'Xiao Hu', OpenAI released the GPT-5.6 update announcement, upgrading the default model for free and Go users to GPT-5.6 Luna, with unlimited free text chat. This strategy continues OpenAI's appeal to free users, but the source is a third-party repost, and the official announcement has not been directly confirmed.
Meanwhile, GPT-5.6 Sol, another version, adjusts conversation style, with more direct answers and streamlined formatting, reducing pleasantries and redundant details. This suggests OpenAI is differentiating model styles, but the specific mechanism is not disclosed.
The announcement claims that in complex tests involving finance, healthcare, law, and other areas requiring precise data, the new model's factual error rate is reduced by 68% compared to GPT-5.5. This figure comes from the reposted content and is a source claim, with no official data or third-party verification yet.
Notably, the testing scope and methodology for this figure are not detailed, so its credibility should be assessed cautiously. User 'Tang' commented in the thread to remind that specific features should be celebrated only after the official page is live.
A new 'Thinking Depth Slider' is added on web and mobile, allowing users to adjust between 'fast mode' for everyday Q&A and deep thinking mode for coding, deep research, or complex decisions. This feature aims to let users control the model's reasoning time.
Free models also get a new 'Think Button' that, when clicked on difficult problems, allows the model to spend more time reasoning to produce higher-quality answers. This may be a differentiated experience OpenAI offers to free users, but specific implementation and effects are not detailed.
The announcement emphasizes safety fine-tuning for users under 18, strictly prohibiting emotional/romantic roleplay and discouraging using AI as a substitute for real interpersonal relationships. It also strengthens filtering and guidance for social risks and extreme content (such as self-harm, binge eating, dangerous activities), guiding minors to seek real-world help when necessary.
This move reflects AI companies' attention to protecting minor users, but specific technical implementation and effectiveness evaluation are not provided.
This news originates from a repost by X user 'Xiao Hu', whose content claims to be OpenAI's announcement but does not provide an official link. User 'Tang' commented that it 'feels like a meeting of product roadmap, user wishes, and marketing copy,' suggesting the content may include speculative elements.
Therefore, all specific figures and feature descriptions are source claims, not confirmed facts. The official announcement's landing, feature details, and actual performance still await confirmation from OpenAI's official page.
This report is based on a third-party X repost; all specific figures and feature descriptions have not been directly confirmed by the official source and are source claims. Readers are advised to rely on OpenAI's official announcement.
The GPT-5.6 update brings default model upgrade, a claimed 68% factual error reduction, thinking depth slider, and enhanced minor protections, but the news originates from a third-party repost, and details await official confirmation.
Primary report
Primary source