We present mass models of the four-image gravitational lens system B1608+656 , based on information obtained through VLBA imaging , VLA monitoring and Hubble Space Telescope ( HST ) WFPC2 and NICMOS imaging .
A mass model for the lens galaxies has been determined that reproduces ( i ) all image positions within the observational errors , ( ii ) two out of three flux-density ratios within about 10 % from the observed ratios and ( iii ) the model time delays within 1 % from their observed values , given our best estimate of the Hubble parameter .

Using the time delays determined by Fassnacht et al .
( 1999a ) , we also find that the best isothermal mass model gives H _ { 0 } = 59 ^ { +7 } _ { -6 } km s ^ { -1 } Mpc ^ { -1 } for \Omega _ { m } = 1 and \Omega _ { \Lambda } = 0.0 , or H _ { 0 } = ( 65 - 63 ) ^ { +7 } _ { -6 } km s ^ { -1 } Mpc ^ { -1 } for \Omega _ { m } = 0.3 and \Omega _ { \Lambda } = 0.0–0.7 .
The statistical errors indicate the 95.4 % ( 2– \sigma ) confidence interval .
A systematic error of \pm 15 km s ^ { -1 } Mpc ^ { -1 } is estimated from a 20 % ( 1– \sigma ) uncertainty in the steepness of radial mass profile .

This cosmological determination of H _ { 0 } agrees well with determinations from three other gravitational lens systems ( i.e .
B0218+357 , Q0957+561 and PKS1830-211 ) , Type Ia Supernovae , the Sunyaev-Zel ’ dovich effect and local determinations .
The current agreement on H _ { 0 } – within the 1– \sigma statistical errors – from four out of five gravitational lens systems ( i ) emphasizes the reliability of its determination from isolated gravitational lens systems and ( ii ) suggests that a close-to-isothermal mass profile can describe disk galaxies ( e.g .
B0218+357 and possibly PKS1830-211 ) , ellipticals ( e.g .
B1608+656 ) and central cluster ellipticals ( e.g .
Q0957+561 ) .

The average of H _ { 0 } from B0218+357 , Q0957+561 , B1608+656 and PKS1830-211 , gives H _ { 0 } ^ { GL } = 69 \pm 7 km s ^ { -1 } Mpc ^ { -1 } for a flat universe with \Omega _ { m } = 1 or H _ { 0 } ^ { GL } = 74 \pm 8 km s ^ { -1 } Mpc ^ { -1 } for \Omega _ { m } = 0.3 and \Omega _ { \Lambda } =0.0–0.7 .
When including PG1115+080 , these values decrease to 64 \pm 11 km s ^ { -1 } Mpc ^ { -1 } and 68 \pm 13 km s ^ { -1 } Mpc ^ { -1 } , respectively .
The errors are the estimated 2– \sigma errors on the average .
The Hubble parameter from gravitational lenses seems to agree best with local determinations of H _ { 0 } for a low density universe , under the assumption that all lenses are nearly isothermal .