OpenAI Unveils GPT-6.2: Sol for Productivity, Astra for Excellence
—
# OpenAI reinvents its models with GPT-6.2: two approaches for every need

In the landscape of generative artificial intelligence, OpenAI is tackling a major challenge: delivering exceptional performance without compromising operational efficiency. The company has officially unveiled its new generation of models, dubbed GPT-6.2, structured around two distinct architectures to meet diverse needs: GPT-6.2 Sol for daily productivity and GPT-6.2 Astra for precision-demanding tasks.

GPT-6.2 Sol: The Ideal Productivity Partner
For the majority of users, whether writing code, synthesizing documents, or preparing budgets, OpenAI is betting on the Sol model. It is designed as the perfect balance between execution speed and response quality.
More Natural and Concise Writing
One of Sol’s strengths lies in its writing style. Unlike previous models that could sometimes string sentences together or add unnecessary filler, Sol adopts a more direct and concise tone. This evolution is particularly appreciated when drafting technical or professional content where clarity is paramount.
Unmatched Performance for Daily Tasks
OpenAI has placed significant emphasis on realistic benchmarks. GPT-6.2 Sol now outperforms GPT-6 Astra on the **LifeVal** test, specifically designed to evaluate AI on concrete daily tasks. The economic argument is also strong: this model costs seven times less while offering better performance for these common use cases.
Competitive Pricing
For those who wish to integrate Sol into their workflows via the API, the rates are attractive:
* **Inputs**: $2 per million tokens.
* **Outputs**: $10 per million tokens.
* **Cache**: A reduced cost of $0.10 per million tokens.
GPT-6.2 Astra: The Benchmark for Absolute Quality
While Sol aims for efficiency, Astra positions itself as the frontier model dedicated to projects where perfection is non-negotiable. This is where the new model’s strength shines.
Visual Mastery and Source Fidelity
One of Astra’s major assets concerns presentation generation. The model excels at faithfully reproducing layouts, color palettes, and typography provided in the source. Whether the input is a screenshot or an original file, Astra manages to transform these visual references into coherent slides.
Technical Excellence and Complex Logic
Beyond visuals, Astra’s software capabilities have been strengthened. It demonstrates increased precision for 3D design, complex simulations, and spreadsheet manipulation. The model is capable of performing advanced modifications in specialized software with a rigor that Sol is not necessarily prioritized to provide.
Slightly Higher Cost but Justified
Although more expensive than Sol, Astra remains competitive in the high-performance models market. API rates are set at:
* **Inputs**: $8 per million tokens.
* **Outputs**: $40 per million tokens.
These prices represent a drop of about 20% compared to previous generations, making this power accessible.
Immediate Availability for All
The GPT-6.2 family is now available starting today. Users with a Plus, Pro, Enterprise, or Edu subscription can already test these models in ChatGPT Work and Codex. Additionally, developers have access to these new APIs to integrate these advanced capabilities directly into their applications.
With this dual approach, OpenAI confirms its will to dominate the market by covering both mass needs and those of demanding experts.
—