Basis builds AI agents to automate manual accounting work, helping accountants shift from repetitive tasks to strategic work. With GPT-6 Astra, the company sees a stronger understanding of user intent. Co-founder Mitch Troyanovsky says, "GPT-6 Astra does a better job of really understanding the intent of the user and the problem."

Basis compared GPT-6 Astra and GPT-5.6 Sol on a complicated tax workbook with 50 tabs. GPT-6 Astra completed the workbook in half the time. Troyanovsky notes, "GPT-6 Astra is able to complete that workbook in half the time that GPT-5.6 Sol is able to." The model also makes better decisions at the start, reducing time spent correcting mistakes and improving token efficiency.

GPT-6 Astra adjusts reasoning effort dynamically, dialing up computation for difficult steps and using less for easier ones, while keeping its cache intact. This reduces cost and response time for long tasks, making them more economical for Basis and its customers.

Basis saw about a 20% improvement in internal evaluation scores with GPT-6 Astra, driven by better user intent understanding, including when to ask questions and flag assumptions. The model infers expectations from broader context, reducing the need for explicit instructions and boosting confidence in handling real-world situations beyond internal tests.