OpenAI has introduced GPT-6.2 Astra and GPT-6.2 Sol, two models designed for complex projects and everyday work, respectively. Both models will be available through ChatGPT Work and Codex, as well as via the API.
GPT-6.2 Astra is the company’s advanced model for its most demanding projects. OpenAI says its slides, spreadsheets and documents look less like AI-generated content, while its code is cleaner and the websites and games it creates look and perform better.
On Agents’ Last Exam, a benchmark designed to evaluate AI agents’ ability to use software in professional workflows, GPT-6.2 Astra scored 77.6% at the maximum effort level, at a cost of $4.47. For comparison, GPT-6 Astra, introduced in September 2026, scored 73.1% on the same benchmark at a cost of $8.99.
OpenAI also considers GPT-6.2 Astra the world’s best model for scientists and the company’s most capable model to date for cybersecurity-related tasks.
GPT-6.2 Sol is a general-purpose model designed for everyday work. Compared with GPT-6.1 Sol, introduced in late September, the new version is smarter and more efficient, allowing users to complete more tasks within the same subscription plan. OpenAI said both GPT-6.2 models were developed with the company’s strictest safety measures.
GPT-6.2 Astra and GPT-6.2 Sol are available in ChatGPT Work and Codex to Plus, Pro, Enterprise and Edu subscribers. Users on the $500-per-month Pro 500 plan can access Ultrafast mode with either model. Subscribers to the $200-per-month Pro 200 plan will initially be able to use Ultrafast mode only with GPT-6.2 Sol.
The standard API pricing is as follows:
ModelInput tokens (per 1 million)Cached input tokens (per 1 million)Output tokens (per 1 million)
GPT-6.2 Astra$8$0.40$40
GPT-6.2 Sol$2$0.10$10
