TransBench
Get Featured
Independent measurement bench

We measure translation quality.
Vendors don't get to.

Every machine-translation API below is scored by running the same public FLORES-200 sentence slice through it, measuring latency, and computing a COMET quality score ourselves. The numbers are first-party measurements — and sponsors cannot edit them.

Method: FLORES-200 dev-test (CC BY-SA 4.0) · quality = COMET (wmt22-comet-da) rescaled 0–100, mean over a documented language-pair slice, with sacrebleu BLEU as secondary · latency = median of 5 runs · v1-illustrative

⚠ The numbers below are ILLUSTRATIVE samples. The method is real and reproducible — replace these with measured numbers after running the benchmark with real API accounts. No real accounts, keys, or money were used.

83.0
1200ms
8 tools
Try

Get your translation API featured on TransBench

Featured placement puts your service at the top of the index with your logo and pitch — in front of teams actively choosing a translation API.

Claim or feature your listing →

Placement is for sale. Your measured COMET score is not — every featured service shows the same unedited benchmark number.