OpenAI Whisper — Mongolian Speech-to-Text Benchmark
Real results from the OpenAI Whisper Batch API (enhanced operating point, language mn) evaluated across 4 Mongolian datasets and 265 audio samples. Every number below is measured — not marketing. WER, speed, and pricing are shown as-is, brutally honest.
265/265 samples transcribed · 100% success rate
Lower is better · across 265 samples
Pricing
OpenAI Whisper list pricing for batch transcription. No discounts, no negotiated rates applied — the raw per-minute rate.
Pricing source: OpenAI Whisper public pricing. Duudlaga Flow is shown for context only — this page isolates OpenAI Whisper so the number is not padded by our own product.
WER & CER by dataset
Word and character error rates per dataset. Lower is better — and these are the real OpenAI Whisper numbers, which are weak on Common Voice 24.
Dataset summary
Aggregate accuracy, speed, and timing for each dataset.
| Dataset | Source | Samples | Success | WER | CER | Accuracy | Speed | Audio (s) | Proc (s) |
|---|---|---|---|---|---|---|---|---|---|
| Common Voice 24 (MN) | Hugging Face → | 59 | 59/59 | 104% | 63.4% | -4.0% | 1.95× | 330.6s | 169.5s |
| Shunya Labs Mongolian Speech | Hugging Face → | 60 | 60/60 | 104.1% | 61.3% | -4.1% | 1.39× | 645.1s | 462.7s |
| Common Voice 20 (MN) | Hugging Face → | 54 | 54/54 | 111.6% | 69.7% | -11.6% | 1.96× | 269.8s | 137.6s |
| Modern Voice | — | 92 | 92/92 | 103% | 61.9% | -3.0% | 1.3× | 603.8s | 464.9s |
WER vs speed — per sample
Each dot is one audio sample. The sweet spot is the bottom-left (low error, fast). OpenAI Whisper sits high on error for many Common Voice 24 samples.
Per-sample results
Ground truth shown verbatim in the Expected column. The result column highlights only the words OpenAI Whisper got wrong, in red — no strikethrough/swap gymnastics, just the mistakes.
| # | Audio | Sample | Dataset | Expected (ground truth) | OpenAI Whisper result | WER | CER | I/D/S |
|---|---|---|---|---|---|---|---|---|
| 1 | btsee_0001 | Common Voice 24 (MN) | Гэхдээ амьсгал хураахаасаа өмнө танд мэдэж байгаагаа хэлье. | Ұлдэ әм сұл ұраңаса омін тәнді мидді джаға ға хелі. | 125.0% | 55.2% | 2/0/8 | |
| 2 | btsee_0002 | Common Voice 24 (MN) | Надад заяасан аз жаргал гэдэг ердөө гуравхан сарын хугацаатай байсан гэж үү? | Нады дзэйэ санғад джырғал кетегерді Ұрғун сарай ұқыцат табай сіңгіджу. | 100.0% | 58.7% | 0/2/10 | |
| 3 | btsee_0003 | Common Voice 24 (MN) | Одоо бид өөрсдөө өвчин эмгэгээсээ салахыг хичээцгээе. | Ұтағди дұрс дұршиң үйгісі салғих джүцүгі. | 100.0% | 73.1% | 0/1/6 | |
| 4 | btsee_0004 | Common Voice 24 (MN) | Би бол голдуу хээрээр гэр, хэцээр дэр хийж явдаг хүн. | Би бас қалту хэлээр кир үсгээр дирхи джақты қун. | 90.0% | 49.0% | 0/1/8 | |
| 5 | btsee_0005 | Common Voice 24 (MN) | Хан хурмаст уурлаж, Болдоггүй Бор өвгөнийг хор луугаараа ниргүүлэхээр явуулжээ. | Қан құрымас дұғардас бұл түкө бұрауңы ұңы қоҵірд лоңғар нергүр шер ең бұл дзең. | 140.0% | 71.4% | 4/0/10 | |
| 6 | btsee_0006 | Common Voice 24 (MN) | Алив наашаа ороод ир гээд гэртээ оров. | Әлімн әше орадыр, кеткісті орығы. | 100.0% | 67.6% | 0/2/5 | |
| 7 | btsee_0007 | Common Voice 24 (MN) | Өө өндөр дээдэс таны тухайд би баталж чадахгүй. | Ө, өндөрді цітаңи тұқад, бі бақталік жәл тұқғи. | 100.0% | 56.5% | 0/0/8 | |
| 8 | btsee_0008 | Common Voice 24 (MN) | Харин гурав дахь удаагаас эхлэн хүмүүсийг сонирхож эхлэв. | Қырын үріптүң ұтаңыз екілің Ұмұс ек сәңұрғы джіңіліу. | 100.0% | 76.8% | 0/0/8 | |
| 9 | btsee_0009 | Common Voice 24 (MN) | Та нар очингуутаа шөл л өгч үз. | Таныр ұчыңғута шүлілд ұқчүд. | 100.0% | 50.0% | 0/3/4 | |
| 10 | btsee_0010 | Common Voice 24 (MN) | Ерөөсөө литр үйлдвэрээсээ салаагүй явсан юм чинь. | Еру сөл өлтур өссес әлта өгі өп сөйімчін. | 114.3% | 70.8% | 1/0/7 |
Methodology
How these numbers were produced.
Provider: OpenAI Whisper (OpenAI whisper-1 hosted transcription (auto language detection; 'mn' is not an accepted language hint).).
Endpoint https://api.openai.com/v1/audio/transcriptions. Language mn. Operating point whisper-1. Diarization none.
Datasets: Common Voice 24 (MN), Shunya Labs Mongolian Speech, Common Voice 20 (MN), Modern Voice — 265 samples, 1849.3s of audio total.
- Common Voice 24 (MN)https://huggingface.co/datasets/btsee/common-voices-24-mn
- Shunya Labs Mongolian Speechhttps://huggingface.co/datasets/shunyalabs/mongolian-speech-dataset
- Common Voice 20 (MN)https://huggingface.co/datasets/warmestman/common-voice-20-mn-normalized
- Modern Voice—
Metrics: WER and CER are computed with a standard word/character Levenshtein alignment, normalized for case and punctuation. Accuracy = 100 − WER. Speed factor = audio duration ÷ processing time (× realtime). All requests are real OpenAI Whisper Batch API calls, not cached or simulated.
Diff highlighting: The result column aligns to the ground truth and colors every substitution and insertion red. Deletions (words missing from the result) are not shown in the result column — the Expected column already holds the full ground truth as-is.