The ranker · benchmark 06 / 2026

Objective transcription benchmarks.

Every model. Same audio. Same metrics. Re-run on every model or pricing change.

Start transcribing Open playground

Models Ranked

932

Total Benchmarks

Languages Tested

Jun 12, 2026

Last Updated

01 · Leader

Ink-Whisper

Cartesia

81.3

score

WER

22.80%

Latency

947ms

Cost

$0.0022/min

02 · Runner-up

GPT-4o Mini Transcribe

OpenAI

78.1

score

WER

20.32%

Latency

1.9s

Cost

$0.003/min

03 · Third

Nova-2 Phone Call

Deepgram

76.4

score

WER

17.93%

Latency

1.0s

Cost

$0.0058/min

Language

Category

Accent

Overall

Accuracy

Cheapest

Fastest

#	Model	Overall Score	WER	CER	Latency	Cost	Benchmarks
01	Ink-Whisper Cartesia	81.3	22.80%	15.08%	947ms	$0.0022/min	35
02	GPT-4o Mini Transcribe OpenAI	78.1	20.32%	13.73%	1.9s	$0.003/min	35
03	Nova-2 Phone Call Deepgram	76.4	17.93%	12.44%	1.0s	$0.0058/min	25
04	Nova-2 Voicemail Deepgram	75.7	17.99%	12.35%	1.2s	$0.0058/min	25
05	Nova-2 Meeting Deepgram	73.9	17.22%	11.78%	2.0s	$0.0058/min	25
06	Nova-2 Conversational AI Deepgram	72.2	19.10%	13.08%	2.2s	$0.0058/min	25
07	Nova-2 Finance Deepgram	71.5	19.39%	13.41%	2.4s	$0.0058/min	25
08	Nova-3 Deepgram	69.7	22.03%	14.88%	1.3s	$0.0077/min	34
09	Scribe v2 ElevenLabs	68.5	22.74%	27.14%	4.0s	$0.004/min	35
10	Nova-2 Deepgram	67.8	31.44%	41.14%	1.6s	$0.0058/min	34
11	Universal-3 Pro AssemblyAI	67.1	17.30%	9.93%	5.8s	$0.0035/min	35
12	Whisper 1 (API) OpenAI	65.1	24.09%	13.61%	3.6s	$0.006/min	35
13	GPT-4o Transcribe OpenAI	63.1	33.37%	26.27%	2.8s	$0.006/min	35
14	Whisper Large V3 OpenAI	63.0	24.07%	13.64%	4.3s	$0.006/min	35
15	Enhanced Deepgram	61.7	25.24%	15.53%	1.9s	$0.0165/min	31
16	Speechmatics Standard Speechmatics	59.2	19.05%	11.31%	7.1s	$0.005/min	34
17	AssemblyAI Best AssemblyAI	57.5	20.48%	11.80%	4.1s	$0.09/min	35
18	Speechmatics Enhanced Speechmatics	49.6	15.60%	8.63%	8.7s	$0.0083/min	34
19	GPT-4o Transcribe Diarize OpenAI	49.1	17.85%	11.86%	18.1s	$0.006/min	33
20	Base Deepgram	48.8	56.14%	72.41%	1.1s	$0.0145/min	34
21	Amazon Transcribe Amazon Web Services	47.7	20.63%	11.85%	11.9s	$0.006/min	35
22	Chirp 3 Google Cloud	43.0	14.05%	8.65%	10.3s	$0.0107/min	34
23	Amazon Transcribe Medical Amazon Web Services	42.1	15.86%	11.08%	11.4s	$0.075/min	25
24	Google Telephony Google Cloud	41.8	16.44%	10.52%	22.3s	$0.016/min	29
25	Google Latest (Long) Google Cloud	36.3	27.50%	17.65%	13.5s	$0.0107/min	34
26	Google Command & Search Google Cloud	30.3	39.37%	28.27%	11.6s	$0.016/min	34
27	Google Default Google Cloud	30.0	39.90%	28.43%	11.9s	$0.016/min	34
28	Google Latest (Short) Google Cloud	21.2	57.54%	51.69%	12.1s	$0.016/min	68

Ink-Whisper

81.3

22.80% WER

947ms

$0.0022/min

GPT-4o Mini Transcribe

78.1

20.32% WER

1.9s

$0.003/min

Nova-2 Phone Call

76.4

17.93% WER

1.0s

$0.0058/min

Nova-2 Voicemail

75.7

17.99% WER

1.2s

$0.0058/min

Nova-2 Meeting

73.9

17.22% WER

2.0s

$0.0058/min

Nova-2 Conversational AI

72.2

19.10% WER

2.2s

$0.0058/min

Nova-2 Finance

71.5

19.39% WER

2.4s

$0.0058/min

Nova-3

69.7

22.03% WER

1.3s

$0.0077/min

Scribe v2

68.5

22.74% WER

4.0s

$0.004/min

Nova-2

67.8

31.44% WER

1.6s

$0.0058/min

Universal-3 Pro

67.1

17.30% WER

5.8s

$0.0035/min

Whisper 1 (API)

65.1

24.09% WER

3.6s

$0.006/min

GPT-4o Transcribe

63.1

33.37% WER

2.8s

$0.006/min

Whisper Large V3

63.0

24.07% WER

4.3s

$0.006/min

Enhanced

61.7

25.24% WER

1.9s

$0.0165/min

Speechmatics Standard

59.2

19.05% WER

7.1s

$0.005/min

AssemblyAI Best

57.5

20.48% WER

4.1s

$0.09/min

Speechmatics Enhanced

49.6

15.60% WER

8.7s

$0.0083/min

GPT-4o Transcribe Diarize

49.1

17.85% WER

18.1s

$0.006/min

Base

48.8

56.14% WER

1.1s

$0.0145/min

Amazon Transcribe

47.7

20.63% WER

11.9s

$0.006/min

Chirp 3

43.0

14.05% WER

10.3s

$0.0107/min

Amazon Transcribe Medical

42.1

15.86% WER

11.4s

$0.075/min

Google Telephony

41.8

16.44% WER

22.3s

$0.016/min

Google Latest (Long)

36.3

27.50% WER

13.5s

$0.0107/min

Google Command & Search

30.3

39.37% WER

11.6s

$0.016/min

Google Default

30.0

39.90% WER

11.9s

$0.016/min

Google Latest (Short)

21.2

57.54% WER

12.1s

$0.016/min

Accuracy tradeoffs

WER vs. cost & speed

Category leaders

no single winner

Code-Switching

Scribe v2

ElevenLabs

15.81% WER

Conversational

Chirp 3

Google Cloud

6.36% WER

Finance

GPT-4o Transcribe Diarize

OpenAI

6.94% WER

General

Amazon Transcribe Medical

Amazon Web Services

4.54% WER

Legal

Universal-3 Pro

AssemblyAI

3.20% WER

Medical

Chirp 3

Google Cloud

1.94% WER

Noisy Environment

Nova-3

Deepgram

21.95% WER

Technical

Scribe v2

ElevenLabs

2.81% WER

How the score is built

weighted composite

50%

Accuracy

WER + CER vs. reference transcripts

30%

Speed

Median end-to-end latency

20%

Cost efficiency

Price per minute of audio

How We Benchmark

open methodology

Golden set

Curated test audio with verified reference transcripts across languages, accents, and noise levels

Same audio, every model

Every provider runs the identical test set. No cherry-picked clips, no provider-tuned inputs.

Event-driven & automated

Benchmarks re-run automatically on every model or pricing change, with no human bias. Every provider gets the same test audio.

Scoring

Overall score is a weighted composite: 50% accuracy (WER), 30% speed, 20% cost efficiency

Evaluation metrics

how errors are counted

WER

Word Error Rate

(Wrong + Extra + Missed words) ÷ Total words spoken

Percentage of words incorrectly transcribed (lower is better)

CER

Character Error Rate

(Wrong + Extra + Missed characters) ÷ Total characters spoken

Percentage of characters incorrectly transcribed (lower is better)

MER

Match Error Rate

(Wrong + Extra + Missed) ÷ (Correct + Wrong + Extra + Missed)

Ratio of errors to total alignment length (lower is better)

WIL

Word Information Lost

1 − (Correct ÷ Words spoken × Correct ÷ Words predicted)

Fraction of word information lost in transcription (lower is better)

Route to whichever model wins.

One endpoint, every provider. Pin the leader or let us auto-route to the best model under your accuracy and latency budget.

Get started for free Read the spec ↗