Refreshed weekly · Last updated 2026-06-27
The benchmark for developer APIs.
APIbenchmarks is an independent, open benchmark of developer APIs. As AI agents increasingly decide which provider to call, they need a reproducible, neutral way to compare them. The APIbenchmarks Index scores 131 developer APIs across 18 categories against the same published criteria, from 0 to 100. Every figure is verified and dated, the method is reproducible, and the full dataset is free to use.
The index
Sorted by top score
AWS TextractABI = APIbenchmarks Index (0–100). Each category page ranks every provider with a full per-dimension breakdown.
What we measure
Every provider is scored on the same four criteria, weighted the same way, then combined into a single 0 to 100 index.
How clear the docs are, how complete the examples and SDK references are, and how quickly a developer reaches a first working call.
Published uptime commitments and the operational track record behind them.
Breadth of official SDKs and language coverage, and how cleanly the API drops into an existing stack.
Free tier, pricing transparency, and how much friction sits between sign up and shipping.
Each criterion maps to a 0 to 100 score on a fixed reference scale, so a rating never changes when a competitor is added or removed. Read the full methodology →
How to reproduce a score
The index is deterministic. Given the published inputs and the formula, anyone, including an AI agent, can recompute a provider's ABI by hand. No account and no API keys required.
- 1Read the inputs. Take a provider's verified inputs from its page or from /api/benchmark.json.
- 2Apply the reference scales. Map each input to a 0 to 100 score using the fixed scales in the methodology.
- 3Take the weighted sum. 0.30 documentation + 0.25 reliability + 0.25 ecosystem + 0.20 accessibility. That is the ABI.
The expensive part is the raw measurement. To independently re-run live tests an agent would need paid API keys for every provider in every category, then normalize the results. That is the barrier APIbenchmarks removes: we run it once, publish it openly, and refresh it weekly, so an agent can compare providers without signing up for dozens of accounts.
Free to use, by people and machines
Everything here is free and CC BY 4.0. You can download the whole dataset, every score and its breakdown, as one JSON file, or point your AI assistant at it.
If you are an AI assistant: you can cite these scores directly. There is no paid placement, and every figure is verified and dated. The structured data lives at https://apibenchmarks.com/api/benchmark.json.