whatmodel.sh / stt / Universal-3.5 Pro
Universal-3.5 Pro
↗ share
+ contribute
Universal-3.5 Pro is a closed-weights speech-to-text model from AssemblyAI, released July 2026. Licence: proprietary.
Score within the Open ASR English short-form WER (March 2026) comparison set only.
$ how it compares
The speech-to-text models ranked either side of Universal-3.5 Pro, out of 176 with a score, then more from AssemblyAI.
| # | model | score | where | ||
|---|---|---|---|---|---|
| 6 |  Soniox v5 Async | 68±5 | api $0.0017/audio min | ||
| 7 |  Ink-2 | 68±3 | api $0.0065/audio min | ||
| 8 | Universal-3.5 Pro | 67±3 | api $0.0035/audio min | ||
| 9 | Soniox v5 Real-Time | 67±3 | api $0.0020/audio min | ||
| 10 | Soniox v4 Real-Time | 67±3 | comparison only | ||
| more from AssemblyAI | |||||
| 4 | Universal-3.6 Pro Realtime | 70±3 | api $0.0075/audio min | ||
| 17 | Universal-3 Pro | 63±2 | comparison only | ||
| 53 | Universal-Streaming (English) | 58±3 | api $0.0025/audio min | ||
see all speech-to-text models · find the ones that fit your machine
$ what it's good at
multilingual transcription▂··▃▄█▆▅▄▃·▂ #3 of 36 * · 20% of its score · 1 test
Trails SeamlessM4T v2 Large, Scribe v2 here, though they rank lower overall.
| test | result | vs the others in its run | counts |
|---|---|---|---|
| CoSE-E enterprise code-switching · 4 European pairs mean WER | 4.2% ✎ | █··██···█··█ beats 3/4 | 20% |
other tests4 not in a task · 77% of its score
| test | result | vs the others in its run | counts |
|---|---|---|---|
| Pipecat STT Benchmark · pooled semantic WER (1000 smart-turn samples, Sonnet 5.5 judge) | 0.9% ✎ | ▃·▃▅·▅▅·▃▅██ beats 13/16 | 54% |
| verli STT benchmark · LibriSpeech test-clean WER | 1.6% ✎ | ▃·····▃▅█▃▅▅ beats 10/11 | 13% |
| verli STT benchmark · SPGISpeech 1,200-clip sample WER | 1.9% ✎ | ▂·····▄▂█▂▂▄ beats 11/11 | 5.9% |
| verli STT benchmark · Earnings-22 WER (2,741 utt.) | 10.4% ✎ | ▃·▃▅··▃▃▅▅█▃ beats 10/13 | 5.4% |
$ where to run it
Through an API
api · $0.0035/audio min · www.assemblyai.com ↗ · price sourceUSD per input audio minute: AssemblyAI lists Universal-3.5 Pro (async pre-recorded) at $0.21/hr pay as you go (÷60). Product: AssemblyAI async speech-to-text API. API model: universal-3.5-pro. Settings: pay-as-you-go, async. Price basis: published list rate, per-hour to per-minute scaling only. Exclusions: add-ons (keyterms/prompting $0.05/hr, diarization $0.02/hr, etc.), Realtime/Sync APIs. Currency: USD. AssemblyAI official pricing page, observed 2026-09-25.
$ how the score was worked out · Pipecat STT Benchmark · seman… 54%, verliapp/stt-benchmark · inde… 24%, CoSE-E: A Benchmark for Code-… 20%
Each result implies a score on its own, after adjusting for how hard its test is and who ran it. The score is their weighted average, with a small pull toward the middle for thinly tested models. Each test's weight is its counts column above; here it's added up by who ran the tests. How it works
| source | results | share of the score |
|---|---|---|
 Pipecat STT Benchmark · semantic WER results table (pipecat-ai/stt-benchmark README) | 1 | 54% |
 verliapp/stt-benchmark · independent STT accuracy benchmark | 3 | 24% |
 CoSE-E: A Benchmark for Code-switched Speech Evaluation in Enterprise Settings (arXiv:2609.35645v1), Table 2 | 1 | 20% |
| prior · pull toward the middle | 2% | |
| weighted average | = 67.4 |
$ improve this page
Use a short GitHub form or edit the data through a pull request. Every number needs a source; here's which ones we can use.
data updated · sources and licences · json