We determine and compare technical efficiency (TE), technology gap ratios (TGRs) and meta-frontier technical efficiency (MTEs) of maize production between regions using nationally representative panel data collected from 4001 smallholder farm households in Zambia. We estimate the stochastic meta-frontier and region-specific stochastic frontiers based on the ‘true random effects’ framework. Our results show variations in efficiency measures and that smallholder maize production is characterised by increasing returns to scale across all regions, which clearly suggest maize farmers to reduce their average long-term costs by increasing their production scale. We find that some regions are on average more technically efficient than others while those with TE values exceeding 90 per cent operate further below their potential output than those with moderate TE values. Similarly, farm households from regions whose mean TE values are about 90 per cent employ inferior farming techniques to those employed by farmers from regions whose mean TE values are lower. This is in part due to industry-wide specific environmental factors. Most importantly, we find no region to have maize farmers that adopt the most advanced techniques. Results further indicate that all provinces have had either lower or higher TEs, TGRs and MTEs in one period than in another. Generally, our results point to the need to promote superior techniques that would withstand industry-wide specific environmental factors. While it is not possible to find the many reasons for wide variations in TEs, TGRs, and MTEs across regions and time, our results make novel contributions to literature.