2 models compared on the same benchmark — accuracy, latency, price, language coverage and capabilities. Azure Speech leads on raw accuracy; Azure Speech is fastest; Nova-3 covers the most languages; Azure Speech is cheapest.
Verdict · who wins what
Most accurate
Azure Speech
15.3% WER
Microsoft Azure
Fastest
Azure Speech
0.3× realtime
Microsoft Azure
Most languages
Nova-3
47 languages
Deepgram
Cheapest
Azure Speech
$0.0060/min
Microsoft Azure
Comparing
2 models · same benchmark
1
Azure Speech
Microsoft Azure
Azure default STT model — 140+ languages, diarization, word timestamps
2
Nova-3
Deepgram
Deepgram's flagship model — 53% lower WER vs competitors, code-switching support
general
legal
medical
noisy
technical
uk
us
Latency p99
2.0s
1.8s
Final drain
0ms
181ms
Real-time factor
1.01×
1.01×
Flicker
13.2%
9.6%
Cadence
1.7/s
0.8/s
$36.00
$46.08
1,000 hours
$360.00
$460.80
Yes
Auto-detect language
No
Yes
Live / streaming
Yes
Yes
Custom vocabulary
No
Yes
Max file size
1 GB
2 GB
Max duration
4 h
No limit
Compliance
Is Azure Speech or Nova-3 more accurate?
Azure Speech is more accurate, with a 15.3% word error rate versus 21.0% for Nova-3, on our standardized benchmark.
Which is cheaper, Azure Speech or Nova-3?
Azure Speech is cheaper at $0.0060/min versus $0.0077/min for Nova-3.
Should you choose Azure Speech or Nova-3?
Choose Azure Speech for accuracy, speed, and lower cost; choose Nova-3 for language coverage.
