GPT-6.2-Astra Cuts Medical Computer-Use Task Costs Nearly 5x in Real-World Tests
Prescience's internal A/B tests on real patient medical computer-use tasks show GPT-6.2-Astra cuts average task calls to about 1/1.7, with fewer trial-and-error steps and roughly 4.9x lower overall cost.
2026-11-22 ~ 2026-11-22 · 2 related posts
- GPT-6.2-Astra cuts healthcare computer-use agent costs 4.9x in Prescience evals — RishabJainK · 2026-11-22
1 near-duplicate retellings: RishabJainK