In principle, it's done like so. Encode to lower bitrate with no audio. Of course, this assumes you have H.264 video. Check the encoding guide for more info.
ffmpeg -i input.mp4 -c:v libx264 -b:v 700k -an lower.mp4
Then, merge your files:
ffmpeg -i input.mp4 -i lower.mp4 -c copy -map 0:v -map 0:a -map 1:v output.mp4
This will take the first file's first video and audio streams, and the second file's first video stream, and copy them to the output container.
Then again, as @trikly mentions, there's no real advantage in doing this. Especially since you've tagged your question with "video streaming", note that by adding yet another stream inside a container, you're increasing the size of the file that the client has to download, rather than offering just a smaller representation, which is what you typically want to do for restricted bandwidth environments.
The real solution would be to do some kind of adaptive streaming, either with MPEG-DASH or HLS*. ffmpeg can create segmented files that can be used for DASH and HLS streaming, but you also need to generate the meta information files, etc. And you need a client that is DASH or HLS capable.
If that's too much to set up, you should simply offer the video in two different formats, each with their own link.
* Scalable Video Coding was also designed with a similar goal but it's not really used by anybody these days.