Speech-to-Text · Multilingual

Tamil

Language

Request a new language

Tell us which language you want added to this benchmark.

Benchmarks

X
Y
200ms400ms600ms800ms10.0%12.0%14.0%16.0%Finalize latency lower is betterCER lower is betterOpenAI GPT-4o TranscribeDeepgram Nova-3Soniox stt-rt-v5OpenAI GPT-4o-mini Transcribespeko.ai/benchmarks

CER

lower is better

Character error rate, batch path — word scoring mismeasures this language.

GoogleChirp 37.4%
OpenAIGPT-4o TranscribeTry →10.1%
Smallest AISmallest AIPulse ProTry →12.1%
DeepgramNova-3Try →14.2%
Sonioxstt-rt-v5Try →14.6%
OpenAIGPT-4o-mini TranscribeTry →16.1%
CartesiaInk-WhisperTry →30.6%
0.0%10.0%20.0%30.0%40.0%

Finalize latency

lower is better

Time from the caller stopping to the finalized transcript, not the first partial — the agent cannot reply until it lands. Measured on English audio; Tamil has no sweep of its own yet.

DeepgramNova-3Try →189ms · 770ms
Sonioxstt-rt-v5Try →340ms · 1.33s
OpenAIGPT-4o-mini TranscribeTry →822ms · 1.09s
OpenAIGPT-4o TranscribeTry →848ms · 1.18s
0ms300ms600ms900ms1.20s

p50 p90

Time to first token

lower is better

How long before the first word of the transcript arrives. Measured on English audio; Tamil has no sweep of its own yet.

DeepgramNova-3Try →976ms · 1.98s
Sonioxstt-rt-v5Try →1.53s · 2.04s
OpenAIGPT-4o-mini TranscribeTry →6.33s · 10.6s
OpenAIGPT-4o TranscribeTry →6.40s · 10.8s
0ms2.00s4.00s6.00s8.00s

p50 p90

Cost

lower is better

$ per 1,000 minutes of audio · vendor list price, the same in every language