Verified on 10 November 2026. OpenAI facts come from OpenAI's own pages; accuracy figures come from the published accuracy page. Anything not confirmed is left out.
Versions
- GPT-6 Astra was introduced on 3 September 2026. Source: OpenAI release notes.
- GPT-6 Sol and GPT-6 Luna were introduced on 22 September 2026. Source: OpenAI, "Introducing GPT-6 Sol and Luna".
- GPT-6.1 Sol was announced on 29 September 2026. GPT-6.1 Terra began rolling out in ChatGPT Work and Codex on 26 October 2026. Source: OpenAI release notes.
- In the Chat tab, Plus, Pro, Business and Enterprise use GPT-6 Sol, Free and Go use GPT-6 Luna, and the Pro reasoning option uses GPT-6 Astra. Source: OpenAI release notes, 7 and 8 October 2026.
More on each model: ChatGPT versions.
Watermark
- OpenAI announced textGrain on 5 October 2026. It is being added to eligible ChatGPT and Codex text in the EU only. API customers can opt in; it is off by default. Source: OpenAI, "EU text provenance".
- OpenAI said it is not making text watermarking a global default at launch. Same source.
- The text detector is not public. Access is case by case for approved researchers and expert organizations, and checking student work is outside its scope. Source: OpenAI Help Center, "Provenance signals in OpenAI-generated content".
- OpenAI states that the absence of a detected watermark does not prove human authorship. Source: OpenAI, "EU text provenance".
More: ChatGPT watermark status.
What OpenAI says about AI detectors
- OpenAI's educator FAQ says AI detectors do not work in its experience, and that ChatGPT cannot tell whether it wrote a given text. Source: OpenAI Help Center, "How can educators respond to students presenting AI-generated content as their own?".
- OpenAI's own trained detector labeled human text such as Shakespeare and the Declaration of Independence as AI-generated. Same source.
That is why a detector result is a reason to look closer, not proof; see AI detection for teachers.
Veread accuracy
- 0.5% of human-written texts in seven languages wrongly flagged.
- 98.3% of fully AI-written texts detected, and 100% of fully AI-written OpenAI-family texts.
- 95.5% of AI-written and AI-assisted texts of every kind.
- 94.9% on AI models never used in development.
- 92.6% on AI rewrites of human text, and 92.2% on human texts continued by AI.
- Minimum text length: 100 words.
Source: Fileverity's accuracy page, linked above. Run a check on ChatGPTDetect.