wav2vec2 XLSR-53 MN — Mongolian Speech-to-Text Benchmark
Real results from the wav2vec2 XLSR-53 MN 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
wav2vec2 XLSR-53 MN list pricing for batch transcription. No discounts, no negotiated rates applied — the raw per-minute rate.
Pricing source: wav2vec2 XLSR-53 MN public pricing. Duudlaga Flow is shown for context only — this page isolates wav2vec2 XLSR-53 MN 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 wav2vec2 XLSR-53 MN 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 | 37.2% | 11.4% | 62.8% | 37.32× | 330.6s | 8.9s |
| Shunya Labs Mongolian Speech | Hugging Face → | 60 | 60/60 | 44.6% | 15.3% | 55.4% | 45.18× | 645.1s | 14.3s |
| Common Voice 20 (MN) | Hugging Face → | 54 | 54/54 | 31.7% | 9.6% | 68.3% | 37.72× | 269.8s | 7.2s |
| Modern Voice | — | 92 | 92/92 | 58% | 24.3% | 42.0% | 45.1× | 603.8s | 13.4s |
WER vs speed — per sample
Each dot is one audio sample. The sweet spot is the bottom-left (low error, fast). wav2vec2 XLSR-53 MN 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 wav2vec2 XLSR-53 MN got wrong, in red — no strikethrough/swap gymnastics, just the mistakes.
| # | Audio | Sample | Dataset | Expected (ground truth) | wav2vec2 XLSR-53 MN result | WER | CER | I/D/S |
|---|---|---|---|---|---|---|---|---|
| 1 | btsee_0001 | Common Voice 24 (MN) | Гэхдээ амьсгал хураахаасаа өмнө танд мэдэж байгаагаа хэлье. | гэхдээ амьсгал хураахаасаа өмнө танд мэдэж ягаагаа хэлэе | 25.0% | 6.9% | 0/0/2 | |
| 2 | btsee_0002 | Common Voice 24 (MN) | Надад заяасан аз жаргал гэдэг ердөө гуравхан сарын хугацаатай байсан гэж үү? | надад заяасан аз жаргал гэдэг ердөө гуравхан сарын хугацаатай байсан гэж хүү | 8.3% | 1.3% | 0/0/1 | |
| 3 | btsee_0003 | Common Voice 24 (MN) | Одоо бид өөрсдөө өвчин эмгэгээсээ салахыг хичээцгээе. | одоо бид өөрждөө өвчин эмгийдээсээ салахыг хичээцгээ гийдэг | 57.1% | 21.2% | 1/0/3 | |
| 4 | btsee_0004 | Common Voice 24 (MN) | Би бол голдуу хээрээр гэр, хэцээр дэр хийж явдаг хүн. | би бол голдуу хээрээр гэр хэчээр дэр хийж явдаг хүн | 10.0% | 2.0% | 0/0/1 | |
| 5 | btsee_0005 | Common Voice 24 (MN) | Хан хурмаст уурлаж, Болдоггүй Бор өвгөнийг хор луугаараа ниргүүлэхээр явуулжээ. | тав хурамжт уулаж болдоггүй бур вгүн хоёр луугаар нэргүүл хэрэгүлжээ | 90.0% | 29.9% | 0/0/9 | |
| 6 | btsee_0006 | Common Voice 24 (MN) | Алив наашаа ороод ир гээд гэртээ оров. | алив наашаа оройдр гээд гэртээ оров | 28.6% | 8.1% | 0/1/1 | |
| 7 | btsee_0007 | Common Voice 24 (MN) | Өө өндөр дээдэс таны тухайд би баталж чадахгүй. | өө өндөр дээд таны тухайд би баталж чадахгүй | 12.5% | 4.3% | 0/0/1 | |
| 8 | btsee_0008 | Common Voice 24 (MN) | Харин гурав дахь удаагаас эхлэн хүмүүсийг сонирхож эхлэв. | харин гурав даг удаагаас эхлэх хүмүүсийг сонирхож эхлэв | 25.0% | 5.4% | 0/0/2 | |
| 9 | btsee_0009 | Common Voice 24 (MN) | Та нар очингуутаа шөл л өгч үз. | та нар очингуутаа шүлэл өгч үз | 28.6% | 6.7% | 0/1/1 | |
| 10 | btsee_0010 | Common Voice 24 (MN) | Ерөөсөө литр үйлдвэрээсээ салаагүй явсан юм чинь. | ерөөсөө л үйлдвэрээсээ салуугүй орсон юм чинь | 42.9% | 16.7% | 0/0/3 |
Methodology
How these numbers were produced.
Provider: wav2vec2 XLSR-53 MN (wav2vec2 XLSR-53 fine-tuned on Mongolian Common Voice — by download count the most-used Mongolian-specific open ASR model on the Hub. CTC output, so no punctuation or casing.). Models with this badge were trained on data that overlaps the benchmark corpora (Mongolian Common Voice — 113 of 173 samples — or the public Shunya Labs corpus). Scores on overlapping corpora are inflated by memorization; judge these models by the per-dataset table on their details page, especially the corpora they were NOT trained on.
Endpoint local (transformers). Language mn. Operating point anton-l/wav2vec2-large-xlsr-53-mongolian (CTC) on mps. 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 wav2vec2 XLSR-53 MN 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.