Side by side

See the difference.

Scores, capabilities and costs for the models you’re considering.

Scroll to compare models →

Comparison of AssemblyAI: Universal-3.5 Pro, espnet/owsm_ctc_v4_1B
2 modelsassemblyaiAssemblyAI: Universal-3.5 ProRemove ×espnetespnet/owsm_ctc_v4_1BRemove ×
What’s differentUniversal-3.5 Pro is AssemblyAI's speech-to-text model served through its Sync API, returning a complete transcript with word-level timestamps in a single synchronous response for audio up to 20 minutes. It...1B parameters · 75 languages · Encoder: Whisper · Decoder: CTC
DateSep 22, 2026ListedOct 17, 2026Discovered
Listed rate$83.3333 in / $0.00 out/ 1M tokensOpenRouter · all routes ↗See provider quotes
Provider endpoints1 endpoint →0 endpoints →
AMI-Cleaned WER ↗English short-form; normalization unversioned10.63% WER ↗assemblyai/universal-3-5-proLeaderboard entry11.1% WER ↗espnet/owsm_ctc_v4_1BLeaderboard entry
Earnings22-Cleaned-AA-chunked WER ↗English short-form; normalization unversioned6.05% WER ↗assemblyai/universal-3-5-proLeaderboard entry10.31% WER ↗espnet/owsm_ctc_v4_1BLeaderboard entry
Gigaspeech-Cleaned WER ↗English short-form; normalization unversioned7.62% WER ↗assemblyai/universal-3-5-proLeaderboard entry9.29% WER ↗espnet/owsm_ctc_v4_1BLeaderboard entry
LS Clean WER ↗English short-form; normalization unversioned1.1% WER ↗assemblyai/universal-3-5-proLeaderboard entry2.11% WER ↗espnet/owsm_ctc_v4_1BLeaderboard entry
LS Other WER ↗English short-form; normalization unversioned2.11% WER ↗assemblyai/universal-3-5-proLeaderboard entry4.38% WER ↗espnet/owsm_ctc_v4_1BLeaderboard entry
SPGISpeech WER ↗English short-form; normalization unversioned1.81% WER ↗assemblyai/universal-3-5-proLeaderboard entry2.31% WER ↗espnet/owsm_ctc_v4_1BLeaderboard entry
URGENT2024 WER ↗English short-form; normalization unversioned9.08% WER ↗assemblyai/universal-3-5-proLeaderboard entry16.12% WER ↗espnet/owsm_ctc_v4_1BLeaderboard entry
Voice Arena Monsoon WER ↗English short-form; normalization unversioned3.49% WER ↗assemblyai/universal-3-5-proLeaderboard entry6.66% WER ↗espnet/owsm_ctc_v4_1BLeaderboard entry
Voxpopuli-AA-Cleaned WER ↗English short-form; normalization unversioned1.92% WER ↗assemblyai/universal-3-5-proLeaderboard entry4.09% WER ↗espnet/owsm_ctc_v4_1BLeaderboard entry
Context window——
Max output——
Parameters—1B
Audio inputSupportedSupported
TranscriptionSupportedSupported

Add another model

Search all models →