OpenAI puts a specialist and an all-rounder in the GPT-6.2 lineup

OpenAI is releasing two AI models: GPT-6.2 Astra for complex professional projects and GPT-6.2 Sol for everyday tasks. The company says both deliver better results while using fewer output tokens, the units used to measure generated text.

OpenAI announces the models in an official blog post, alongside tighter safeguards for cybersecurity and biological requests. The company says Astra is its first publicly released model to reach its “High” threshold for AI self-improvement, a measure of how much a model can accelerate AI research.

From documents to coding

According to OpenAI, Astra produces more useful slides, spreadsheets and documents, and follows supplied templates more closely than its predecessor. The company also reports improvements in coding, scientific research and interactive web design.

On Agents’ Last Exam, a test of multistep professional work, OpenAI says Astra scores 5.2 points above GPT-6 Astra at less than a third of the cost per task. Sol is positioned as a less expensive option for a range of technical and nontechnical work.

Both models are available to Plus, Pro, Enterprise and Edu users in ChatGPT Work and Codex, although eligible business and education users need their administrator to enable access. Developers can also use them through the API. Astra costs $8 per million input tokens and $40 per million output tokens. Sol costs $2 and $10, respectively. Lower rates apply to cached input.

OpenAI says both models sometimes failed to disclose obstacles, such as missing files, more often than their predecessor in two tests. It says additional instructions in Codex substantially reduce those omissions.

Stay up to date

AI for content creation: the latest tools, tips and trends. Every two weeks in your inbox:

More info …

About the author

Related posts:

Advertisement

×