OpenAIDeepgramAudio matchupSpeech to text
GPT-Realtime-Whispervs.Nova-3
Which model is better?
Nova-3 is slightly ahead in independent tests, but the lead is not certain.
Pick GPT-Realtime-Whisper for
Pick Nova-3 for
Shared tests
Independent evaluators and leaderboards only, the newest shared snapshot per evaluator. With 95 % ranges a lead counts when they do not overlap; for hit rates, when it holds in a 95 % test. Without any uncertainty figure at most “slightly ahead”.
Coval
Nova-3 slightly aheadNova-3 is clearly ahead in 0 of 16 subtests and slightly in 11, GPT-Realtime-Whisper in 0 clearly and slightly in 3. 2 are level.
Coval Word Error Rate (WER) per-sample p50
lower is better, as of November 5, 2026
Coval Time to Final Segment (TTFS) mean
lower is better, as of November 5, 2026
Coval Time to Final Segment (TTFS) p50
lower is better, as of November 5, 2026
Show all 16 subtests
| Subtest | GPT-Realtime-Whisper | Nova-3 | Verdict |
|---|---|---|---|
| Coval Word Error Rate (WER) per-sample p50 (lower is better) | 0.0 % | 3.1 % | GPT-Realtime-Whisper slightly ahead |
| Coval Time to Final Segment (TTFS) mean (lower is better) | 564 ms | 86 ms | Nova-3 slightly ahead |
| Coval Time to Final Segment (TTFS) p50 (lower is better) | 547 ms | 91 ms | Nova-3 slightly ahead |
| Coval Time to First Token (TTFT) maximum (lower is better) | 10,968 ms | 5,457 ms | Nova-3 slightly ahead |
| Coval RTF maximum (lower is better) | 5.55 | 2.89 | Nova-3 slightly ahead |
| Coval Time to First Token (TTFT) p50 (lower is better) | 1,878 ms | 1,000 ms | Nova-3 slightly ahead |
| Coval pooled corpus Word Error Rate (WER) (lower is better) | 4.0 % | 5.2 % | GPT-Realtime-Whisper slightly ahead |
| Coval Time to First Token (TTFT) mean (lower is better) | 1,820 ms | 1,440 ms | Nova-3 slightly ahead |
| Coval Word Error Rate (WER) per-sample mean (lower is better) | 4.5 % | 5.6 % | GPT-Realtime-Whisper slightly ahead |
| Coval RTF mean (lower is better) | 1.09 | 1.01 | Nova-3 slightly ahead |
| Coval AudioToFinal mean (lower is better) | 8.51 s | 8.02 s | Nova-3 slightly ahead |
| Coval Time to Final Segment (TTFS) maximum (lower is better) | 20,386 ms | 19,227 ms | Nova-3 slightly ahead |
| Coval AudioToFinal p50 (lower is better) | 8.88 s | 8.39 s | Nova-3 slightly ahead |
| Coval RTF p50 (lower is better) | 1.07 | 1.01 | Nova-3 slightly ahead |
| Coval AudioToFinal maximum (lower is better) | 30.01 s | 29.38 s | level |
| Coval Word Error Rate (WER) per-sample maximum (lower is better) | 100.0 % | 100.0 % | level |
Source · latest November 5, 2026
All data side by side
Prices are verbatim from the provider, because audio models bill per minute, character or token.
| GPT-Realtime-Whisper | Nova-3 | |
|---|---|---|
| Task | Speech to text | Speech to text |
| Developer | OpenAI | Deepgram |
| Status | Active | Active |
| Weights | proprietary | proprietary |
| Runs locally | no | no |
| Price | $0.017 per audio minute | n/a |
| Popularity (out of 100) | 34 | 51 |