Background: Past research shows that post-traumatic amnesia (PTA) duration is a particularly robust traumatic brain injury (TBI) outcome predictor, but low specificity limits its clinical utility. Objectives: The current study assessed the relationship between PTA duration and probability thresholds for Glasgow Outcome Scale (GOS) levels. Methods: Data were prospectively collected in this multicentre observational study. The cohort was a consecutive sample of rehabilitation patients enrolled in the National Institute on Disability and Rehabilitation Research funded TBI Model Systems (n = 1332) that had documented finite PTA duration greater than 24 h, and 1-year and 2-year GOS. Results: The cohort had proportionally more Good Recovery (44% vs 39%) and less Severe Disability (19% vs 23%) at year 2 than at year 1. Longer PTA resulted in an incremental decline in probability of Good Recovery and a corresponding increase in probability of Severe Disability. When PTA ended within 4 weeks, Severe Disability was unlikely (<15% chance) at year 1, and Good Recovery was the most likely GOS at year 2. When PTA lasted beyond 8 weeks, Good Recovery was highly unlikely (<10% chance) at year 1, and Severe Disability was equal to or more likely than Moderate Disability at year 2. Conclusions: Two PTA durations, 4 weeks and 8 weeks, emerged as particularly salient GOS probability thresholds that may aid prognostication after TBI.