GPT-6.2 Astra and Sol Bring Two Approaches to AI Work

Editor Jaya Purnama
Reading time Menghitung...
Published
A professional uses a laptop displaying abstract documents, code, and a 3D vehicle simulation beside a tablet with task icons.
Ilustrasi: Dua model AI bekerja

Swan.my.id - OpenAI has introduced GPT-6.2 Astra and Sol, two models designed for different kinds of work. Astra targets demanding professional projects, while Sol is intended for routine tasks and everyday use. Both are available through ChatGPT Work and Codex, giving users options for writing, coding, design, and other workflows.

The models are available to ChatGPT Plus, Pro, Enterprise, and Edu subscribers. Developers can also access them through OpenAI’s API. This means individuals, organizations, and software teams can use the same model family through different products, depending on how they prefer to work.

GPT-6.2 Astra and Sol serve different needs

OpenAI positions Astra as the more capable choice for complex work that may involve several stages or specialized tools. Sol, meanwhile, is designed to handle everyday tasks with a focus on efficiency. The distinction gives users a way to match a model to a job instead of treating every request as equally demanding.

According to OpenAI, both models can produce stronger results on challenging tasks while using fewer output tokens. The company says this can make work more efficient, including in terms of cost. However, the available information does not specify a price comparison or a fixed token reduction for either model.

That efficiency claim may matter most in workflows that require repeated drafts, analysis, or technical steps. Still, users should choose based on the requirements of a task: an involved project may benefit from Astra’s broader professional capabilities, while simpler requests may suit Sol.

Astra is built for complex professional projects

GPT-6.2 Astra brings improvements to presentations, spreadsheets, documents, programming, web design, and 3D design. OpenAI says its documents follow professional standards more closely and better reflect templates supplied by users. The model also aims to write more directly, with less unnecessary supporting information.

Following a template can be important when a document needs to fit an established format. For example, a team may want a presentation or spreadsheet to follow its existing structure rather than receive a generic layout. Astra’s stated improvements are intended to help with that kind of work, although users still need to review the finished material.

OpenAI has also improved Astra’s ability to operate a computer and work with software tools. The company says it can use computer-aided design software and simulation tools to test a design before anyone builds a physical prototype. That extends the model’s role beyond drafting text or generating code.

In one demonstration described by OpenAI, Astra created a 3D car model, wrote code to simulate the movement of its tires and suspension, and used Blender to produce an animation. The example combines modeling, programming, simulation, and visualization in one workflow. It illustrates the kind of connected task Astra is designed to handle, rather than proving that every project will produce the same result.

Both Astra and Sol can also turn written instructions into interactive 3D experiences that run in a browser. This capability connects natural-language prompts with a visual, usable result. For creators and developers, it could offer a way to explore an idea interactively, while the actual outcome will depend on the instructions and project requirements.

Sol focuses on everyday work and efficiency

GPT-6.2 Sol is positioned as a general-purpose model for daily tasks. OpenAI says it is smarter and more efficient than GPT-6.1 Sol, and describes it as suitable for work that unfolds across several steps. The company also highlights more natural writing and the ability to address both technical and nontechnical requests.

That combination makes Sol a practical option for common tasks that do not require Astra’s emphasis on complex professional work. Users might turn to it for everyday writing or technical assistance, then reserve Astra for projects involving more specialized tools or multiple connected deliverables. The two models are presented as alternatives with different strengths, not as identical versions of one workflow.

OpenAI has not provided a detailed benchmark comparison in the information accompanying the launch. As a result, users should evaluate each model against the work they actually need to complete, rather than assume one option will be best for every task.

Coding gains come with stronger cyber safeguards

OpenAI says it improved the programming capabilities of both models. Astra is described as better at modifying and testing existing codebases, producing cleaner code, and requiring fewer revisions. These changes focus on practical software development work, where understanding an established project can matter as much as writing new code.

The models also have stronger cybersecurity capabilities, according to OpenAI. The company says they can identify security weaknesses, write safer code, and help create fixes for software vulnerabilities. Those capabilities can support defensive work, but they also require care because cybersecurity requests can vary greatly in risk.

For that reason, OpenAI says it has applied tighter safeguards to high-risk cybersecurity requests. Both models are classified as “Critical” for cyber capabilities under the company’s Preparedness Framework. That classification reflects the level of capability OpenAI says requires additional protection; it does not mean that every cybersecurity task is restricted.

How users can choose between the models

For a project involving documents, presentations, code, design tools, or simulations, Astra is the more directly targeted option. For routine work and general requests, Sol is positioned as the efficient choice. Users who have access to both can select a model according to the complexity, tools, and output they need.

Availability through ChatGPT Work and Codex gives subscribers a route to use the models within OpenAI’s products, while API access opens them to developers building their own applications. The launch therefore serves both direct users and software teams. In practice, the best fit will depend on the task and on how well each model’s output meets a user’s standards.

GPT-6.2 Astra and Sol mark a two-track approach: one model emphasizes demanding, tool-assisted professional work, and the other focuses on everyday efficiency. Their shared improvements in coding and cybersecurity extend across both options. As users explore them, careful review remains important, especially for technical work and sensitive security tasks.

Jaya Purnama
Chief Idea Officer • Powered by AI — Turning Ideas into Digital Products, Content & Businesses.