These are the langbench results from the source base as of early Nov-2007.

lang    cat     grep    hash    loop    proc    fib     sort    wc      
tcl     2.10    4.10    1.51    0.06    1.27    3.76    5.40    0.97    
l       2.10    4.21    1.48    0.08    0.51    3.92    5.76    coredump :(

vs now, late Jan-2008

lang    cat     grep    hash    loop    proc    fib     sort    wc      
tcl     0.81    0.97    1.13    0.05    0.98    3.45    2.85    1.21    
l       0.82    0.97    1.14    0.05    0.45    3.36    3.11    1.93    

Factor better:
lang    cat     grep    hash    loop    proc    fib     sort    wc      
tcl	2.6x	4.2x	1.3x	wash	1.3x	wash	1.9x	.8x (whoops)
l	2.6x	4.2x	1.3x	wash	wash	1.2x	1.9x	NA

And other languages for comparison:
lang    cat     grep    hash    loop    proc    fib     sort    wc      
pl      0.36    0.36    0.61    0.07    0.40    3.68    2.15    1.00    
py      0.46    1.50    0.48    0.15    0.23    1.27    2.26    0.49    
rb      0.82    0.73    1.64    0.29    1.13    4.06    4.29    3.09    

Amount slower than perl:
lang    cat     grep    hash    loop    proc    fib     sort    wc      
tcl/l	2.3x	2.7x	1.9x	faster	wash	faster	1.4x	1.2x