We studied the agreement between plasma clearance of mannitol and the reference method, plasma clearance of 51Cr‐EDTA in outpatients with normal to moderately impaired renal function. Forty‐one patients with a serum creatinine <200 μmol l−1 entered the study. 51Cr‐EDTA clearance was measured with the standard bolus injection technique and glomerular filtration rate (GFR) was calculated by the single‐sample method described by Jacobsson. Mannitol, 0·25 g kg−1 body weight (150 mg ml−1), was infused for 4–14 min and blood samples taken at 1‐, 2‐, 3‐ and 4‐h (n = 24) or 2‐, 3‐, 3·5‐ and 4‐h after infusion (n = 17). Mannitol in serum was measured by an enzymatic method. Plasma clearance for mannitol and its apparent volume of distribution (Vd) were calculated according to Brøchner‐Mortensen. Mean plasma clearance (±SD) for 51Cr‐EDTA was 59·7 ± 18·8 ml min−1. The mean plasma clearance for mannitol ranged between 57·0 ± 20·1 and 61·1 ± 16·7 ml min−1 and Vd was 21·3 ± 6·2% per kg b.w. The between‐method bias ranged between −0·23 and 2·73 ml min−1, the percentage error between 26·7 and 39·5% and the limits of agreement between −14·3/17·2 and −25·3/19·9 ml min−1. The best agreement was seen when three‐ or four‐sample measurements of plasma mannitol were obtained and when sampling started 60 min after injection. Furthermore, accuracy of plasma clearance determinations was 88–96% (P30) and 41–63% (P10) and was highest when three‐ or four‐sample measurements of plasma mannitol were obtained, including the first hour after the bolus dose. We conclude that there is a good agreement between plasma clearances of mannitol and 51Cr‐EDTA for the assessment of GFR.