Original sourceOpenAIAdditional: Amazon Web Services
Summary
OpenAI announced two new models: GPT-6.2 Astra focuses on complex professional tasks, covering presentations, spreadsheets, documents, cleaner code and websites, games, and 3D design frontends; GPT-6.2 Sol is an efficiency-oriented everyday model that is smarter than GPT-6.1 Sol and can handle mult…
Key points
- Developers and enterprises can split work between the two models based on task complexity and re-estimate agentic workflow costs using token efficiency and caching discounts.
- This is the core release of OpenAI's new dual-model strategy, directly changing how enterprises select models and structure costs between professional tasks and everyday automation.
- Teams need to re-evaluate model allocation across Codex and API workflows and watch how Ultrafast and cache pricing affect actual bills for high-frequency agent tasks.
Editorial note
This page is Code & Chain's editorial summary of public sources. It may be prepared with AI assistance and published through an automated workflow. Refer to the original sources; this content is not investment, legal, or tax advice.