I have used Zeplin. It has some great features like allowing to have UI-level discussions threads with developers, allowing to export assets and its integration with slack for communication about updates. It also allows developers to just hover over elements to get dimensions, spacing, type details and color etc. But the developers still need to hover to know these values. Hover on every single element is taxing.
What if the developers just remember the values and the application rules? In that case, they will not need to hover over every single element. They will just look at an element (no need to hover or inspect)—
example 1) determine the hierarchy mentally, know its a header 3 and apply header 3 font-size and line-height
example 2) see it’s a table, know tables take 8 px top and bottom margins for all rows and apply values
example 3) See it’s a blue button, just know the color, dimensions and padding within button and apply values
(Ideally, much of it will also be componentized in a component library)
In that sense, the developers may not need to rely on Zeplin to get to a state of pixel perfection.