customers
All customer stories
Top Voice AI companies are building with Assembly.
resources
Latest Release
Universal 3.5 Pro
Our new flagship speech-to-text model handles real-world audio the way it actually happens. Available for real-time and pre-recorded audio.
resources
Word error rate is easy to game and easy to misread. Better truth files, smarter normalization, and the right metric show what a model actually gets right—then see how Universal-3 holds up.
Q3 revenue came in at $48.2 million, up 19% year over year.
And gross margin expanded to 71% in the quarter.
Right, driven by the Universal-3 rollout to enterprise.
Sloppy reference transcripts inflate error rates. Cleaner, verified truth files reveal what a model actually gets wrong.
Casing, numbers, and punctuation shouldn't count as errors. Normalize before scoring for a fair, apples-to-apples comparison.
WER misses entities, formatting, and readability. Track missed-entity rate and the metrics your users actually feel.
36%
improvement in close rate
“The new Universal-3.5 Pro speech model from AssemblyAI is best so far in terms of accuracy, latency, and language switching.”
80%
increase in customer satisfaction
“Assembly has saved us countless hours managing models, and provided exceptional accuracy.”
36%
improvement in close rate
“The new Universal-3.5 Pro speech model from AssemblyAI is best so far in terms of accuracy, latency, and language switching.”
80%
increase in customer satisfaction
“Assembly has saved us countless hours managing models, and provided exceptional accuracy.”
Create a free account and run AssemblyAI's models against your own files. Compare results in the no-code playground, then wire the API into your eval pipeline.
Start building freeNo credit card required
WER is calculated as (substitutions + insertions + deletions) / total words in the reference transcript. It's the standard metric for evaluating speech-to-text accuracy.
Average normalized WER across 26 real-world datasets.
*Lower is better*Source: AssemblyAI published benchmarks, March 2026.
