Basis compares GPT-6 Astra and GPT-5.6 Sol for accounting workflows

01 What happened

Basis, a provider of AI agents for accounting, released findings from an internal performance test between GPT-6 Astra and GPT-5.6 Sol. According to Basis, the Astra model completed a 50-tab tax workbook in half the time of GPT-5.6 Sol.

02 Key details

• Basis reports that Astra completed the tax workbook in 50% of the time required by GPT-5.6 Sol.
• The firm claims a 20% improvement in internal evaluation scores by better identifying intent and flagging assumptions.
• Basis adjusts reasoning effort throughout task execution, assigning higher computation to difficult steps.
• Evaluations measure agent adherence to document templates and the accuracy of primary source references.

03 Why it matters

The case shows why reasoning effort can be adjusted during a long accounting task and why output quality needs a separate check. These performance figures come from Basis's own tests.

04 Who it matters to

Accounting professionals, AI tool developers and financial analysts.