Teller is a much smaller company and operates mostly in Europe, whereas Plaid is much bigger and operates mostly in the US, so this is really an apples and oranges comparison, especially as the ecosystem of connecting to financial institutions is so different across Europe vs. the US. Europe is heavily regulated and provides more standard ways of obtaining data from financial institutions, the US has much less in that regard.
The Teller team as you allude to is proud of their use of reverse-engineered APIs, although IMO this is still relying on undocumented behavior that can change at any time, is subject to being blocked by banks, etc. Basically, it works until it doesn't.
Plaid has committed to moving 75% of traffic to officially supported APIs by the end of 2021. Plaid hasn't said which banks they support via official API versus which ones they support by other means, although there are a few banks/banking platforms that have announced that they connect to Plaid via API, including Jack Henry (supports over 350 institutions), Bank of America, Capital One, US Bank, and Wells Fargo. (Plaid has also created an open API standard for institutions to use, Plaid Exchange, but hasn't published a list of exactly who's on it aside from Jack Henry.)
Plaid supports around 11,500 supported institutions and the list can be found by creating a developer account and logging in to the dashboard.
Plaid, as a bigger company, offers many features and APIs, as well as SDK and client library support for a number of different platforms and languages.
Plaid currently offers a three-nines overall uptime SLA, but things like robustness and performance are going to vary hugely based on your use case and are super hard to talk about in general terms.
Anyway, if you’re considering Plaid for an app you’re building and have any questions, I’d be happy to chat offline (@Plaiddev is a good way to get in touch if you use Twitter).