← All examples

ModelEthio-ASR Amharic (600M)17Ethio-ASR Multi (300M)19Ethio-ASR Multi (600M)13Ethio-ASR Multi (94M)19Meta MMS-1B19OmniASR-CTC (1B)19OmniASR-CTC (300M)19OmniASR-CTC (3B)19OmniASR-LLM (1B)15OmniASR-LLM (300M)10OmniASR-LLM (3B)10OpenAI Whisper-small19
Audio qualityClean Studio7Ambient Room Noise8Harsh Environment9Distant / Low-Volume Mic6Wideband Phone (VoLTE)8Randomized Phone Channel9Narrowband 2G + Packet Loss13

What did the model get wrong?

Ethio-ASR Multi (600M) Narrowband 2G + Packet Loss   Example 1 of 20  ·  speaker: Female

🎧 Listen — Narrowband 2G + Packet Loss
Heavy background noise plus phone-call compression — worst-case stress test.
✅ What was actually said
ሁለት ዜሮ ሁለት አምስት አንድ ሁለት አንድ ሰባት ሁለት አራት ሰባት ስምንት ዜሮ ሰባት ዘጠኝ አራት
🤖 What this model heard
tir ሽን ሽን ኪን ሁለት ዜሮ ሁለት አምር አን ሁለት አንድንድ ሰባት ሁለትለአራትራ ሰባትባት ስምን ዜሮ ሰባት በራት
Correct
Wrong (different word)
Missed (skipped)
Extra (added)
7
Correct
10
Wrong
2
Missed
1
Extra
37%
Words right
Out of 19 spoken words, this model got 7 right (37%). Errors: 10 words wrong; 2 missed; 1 extra.

Try switching Audio quality — see the same model degrade as the audio gets noisier.