Basis completes tax workbook twice as fast using GPT-6 Astra

By Splitfeed EditorialLast updated Sep 28, 2026, 21:20 UTC

On September 28, 2026, OpenAI published a case study describing how Basis, a tax-focused software company, used GPT-6 Astra to complete a 50-tab tax workbook at twice the speed of GPT-5.6 Sol, the preceding model. Basis attributed the gain partly to GPT-6 Astra's stronger understanding of user intent, and said the result increased its confidence in deploying the model in real-world scenarios. The comparison was run on the same workbook task, making it a direct head-to-head between the two models.

The result matters because tax work is one of the domains where AI deployment carries concrete professional and legal stakes. A 2x speed improvement on a structured, multi-tab workbook is a specific operational claim, not a benchmark abstraction, and Basis framing it as a confidence boost for real-world use signals that the company is moving toward broader deployment rather than controlled testing. What remains in dispute is how well this performance translates beyond a single workbook type, and whether "stronger user intent understanding" is a durable advantage or a narrow fit for this particular workflow.

No reactions from the Pro-AI, Anti-AI, or Middle Ground camps had been published at the time of writing. Pro-AI voices would typically treat a 2x speed gain on professional work as evidence that AI is ready for high-stakes deployment and that productivity arguments for adoption are being validated in practice. Anti-AI voices would likely question what errors or oversights a faster model might introduce in a domain where accuracy carries regulatory consequences. Middle Ground observers would generally call for independent verification of the speed and accuracy claims before drawing conclusions about broader readiness.

The argument would sharpen considerably if Basis publishes error-rate data alongside the speed figures, or if independent researchers test GPT-6 Astra against GPT-5.6 Sol on comparable tax tasks. Any formal accuracy audit tied to real client filings would be the most meaningful signal to watch.