Hi everyone. I'm new to this forum and still getting my bearings with cloud ops, but I've been tasked with researching some academic tools for our research team. We're looking at AI-powered literature analysis platforms.
I need to compare the cost per analyzed document for Iris.ai, Scite, and Dimensions. Our budget is tight, like a carefully sized AWS instance 😅. I'm used to clear pricing for compute, but the pricing pages for these tools are... complex.
Could anyone break down the actual cost per document or per query for these services? For example, in Scite, does one "query" equal one paper? And how does Iris.ai's workspace credit system translate to per-document costs for, say, a systematic review of 1000 PDFs?
Any real-world examples would be a huge help.
Yeah, I felt this exact same way when I started looking at these tools! Coming from a Shopify analytics background, I'm used to clear "cost per X" too.
For Scite, I was told by a colleague that a "query" is indeed a search for a single paper. So checking the citation statements for one article counts as one query. But if you're doing a systematic review of 1,000 PDFs, you'd need to check each one, so that's 1,000 queries. Their pricing plans give you a bundle of queries per month.
I'm actually curious about Iris.ai's workspace credits myself. Their pricing page says one credit can analyze one document, but I think it depends on the document length? Does anyone know if a 50-page PDF costs more than a 10-page one? It seems like it could get expensive fast for a big review.
That colleague is oversimplifying. Scite's definition of a "query" is notoriously flexible. In my experience, a single API call checking multiple citation contexts for one paper can sometimes burn through several query credits. Their pricing page is a masterpiece of obfuscation.
On Iris.ai, the credit cost *does* scale with document size and processing type. A 50-page PDF will absolutely cost more than a 10-pager for any real analysis. The "one credit, one document" line is marketing fluff for the simplest use case. Run a systematic review through it and watch your credits evaporate.
These aren't AWS instances. They're built to be opaque.
Just my two cents.
Opaque pricing is such a pain. Makes it impossible to plan a project budget.
Since it scales with size, does that mean a review with a ton of short conference papers might actually be more manageable cost-wise on Iris.ai than one with fewer, massive articles?