OpenAI released GPT-6.2 Astra and Sol on November 17, 2026.
This follows the release of the GPT-6 family of models, including Astra, Sol, and Luna, as well as Sol 6.1.
OpenAI describes GPT-6.2 Astra as its model for demanding reasoning and coding tasks. Then, it recommends GPT-6.2 Sol as an option for everyday work, including coding and writing.
When new AI models are released, it’s essential for AI detectors to test their efficacy. So, on November 18, we tested whether Originality.ai 5.0 could identify the AI content generated by GPT-6.2 Astra and Sol.
For our study, we tested 1,000 English responses from OpenAI’s GPT-6.2 Astra and Sol with Originality.ai 5.0. These are our findings.
Yes! GPT-6.2 Astra and Sol are highly detectable with Originality.ai.
Originality.ai 5.0 classified 500 of 500 GPT-6.2 Astra responses and 500 of 500 GPT-6.2 Sol responses as AI. Both models had a 100% detection rate in this benchmark.
This study continues our GPT-6 family research.
We started both models with the same prompts so each faced the same mix of writing tasks. A conversational reply calls for a different voice from a news brief, while an expository article gives a model more room to develop an explanation.
Testing five categories lets us look at detection within each type of writing as well as across the full sample.
To build the sample, we sorted each category of our established prompt bank by prompt length and selected every other prompt. That gave us 500 prompts balanced across the five categories, with shorter and longer prompts represented in each.
Astra’s responses ranged from 126 to 2,880 words, and Sol’s ranged from 113 to 3,113. The sample covered short pieces of a few paragraphs and articles long enough to develop a topic in detail.
Originality.ai AI detection identified all of the samples as Likely AI in every writing category, reaching 100% for both models (GPT-6.2 Astra and Sol).
See the table below for results by category:

The key finding from the results noted above? Originality.ai accurately identified 500/500 GPT-6.2 Astra samples and 500/500 GPT-6.2 Sol samples.
Leading AI companies have rapidly released new models throughout 2026, raising questions about whether AI detection tools can still accurately detect generated writing from the latest LLMs.
For editors, students, writers, and teachers, our findings demonstrate that Originality.ai is still a highly accurate way to identify copy generated by the latest LLMs, like GPT-6.2 Astra and Sol.
Try the Originality.ai AI Checker today to maintain transparency in whether writing is human-written or likely AI-generated.
Further Reading:
For the GPT-6 series, this study adds a result for both Astra and Sol at medium reasoning effort, with coverage from short conversational writing to articles of several thousand words. The test measures how often Originality.ai detected known AI writing.
In this study, the result held for all 988 responses from the 494 prompt pairs with identical final wording.
We also checked longer pieces separately. Among responses of at least 300 words, Originality.ai 5.0 classified 421 of 421 Astra responses and 416 of 416 Sol responses as AI.

Math Papers on arXiv are rapidly increasing. In our latest update, Originality.ai found: from October 2025 to October 2026, there was a 115.5% surge in new papers. Further, nearly 1 in 4 sampled abstracts showed signs of probable AI writing.