IIT Madras Hindi ASR Challenge
Speech models transcribe Hindi newspaper passages read aloud. The organizers score word errors and rank open and restricted training separately.
IBM Research and IIT Bombay later evaluated wav2vec 2.0 on the 5.02-hour Hindi test. The original open track ranked eight teams.
- Made by
- Speech Lab, IIT Madras
- Released
- June 2020
- Kind
- Speech recognition and translation
- Level
- Everyday
- Scored
- Checked automatically
- Answers
- Open-ended
- Tasks
- All public
- Measures
- Capability
- Language
- Hindi
- Interaction
- Single turn
- Items from
- Real work and use
- Refresh
- One fixed set
- Run by
- Anyone
- In views
- Audio and speech
ZAPR Media Labs tops the leaderboard with 6.42%
Best score per entry from the benchmark's own leaderboard, read on November 20th, 2026. Where a model ran with several settings or harnesses, we keep its best. Names in gray aren't in our model list: fine-tunes, agent systems, products and older models. They don't count toward leaders.
- 1 ZAPR Media Labs Closed Task 6.42%
- 2 IITH Open Task 7.04%
- 3 SRIB Closed Task 7.08%
- 4 IITB-a Open Task 7.12%
- 5 Vernacular.ai Open Task 7.15%
- 6 IITB-b Closed Task 7.47%
- 7 Armsoftech.air Closed Task 7.71%
- 8 Scribetech Open Task 7.76%
- 9 CDAC-Pune Open Task 8.35%
- 10 IITG Closed Task 10.3%
- 11 IRL-Bangalore Open Task 10.3%
- 12 Zen3 Info Solutions Closed Task 11.9%
- 13 NIT-Kurukshetra Closed Task 37.8%
- 14 IISc Closed Task 42.2%
Source: Official IIT Madras Hindi ASR Challenge leaderboard
| # | Entry | Score | Setting | Submitted by | Date | Source |
|---|---|---|---|---|---|---|
| 1 | ZAPR Media Labs | 6.42% | Closed Task, Kaldi Chain Model, 5-gram SRILM | ZAPR Media Labs | Official leaderboard | |
| 2 | IITH | 7.04% | Open Task, Kaldi Chain Model + Lattice combination of 4-gram lattices | IITH | Official leaderboard | |
| 3 | SRIB | 7.08% | Closed Task, Kaldi Chain TDNN Fusion Model + 4gram LM | SRIB | Official leaderboard | |
| 4 | IITB-a | 7.12% | Open Task, Kaldi Chain Model + LM(ext text) | IITB-a | Official leaderboard | |
| 5 | Vernacular.ai | 7.15% | Open Task, Kaldi Chain Model + RNNLM (ext data) | Vernacular.ai | Official leaderboard | |
| 6 | IITB-b | 7.47% | Closed Task, Kaldi Chain Model + RNNLM | IITB-b | Official leaderboard | |
| 7 | Armsoftech.air | 7.71% | Closed Task, Kaldi Chain Model 4 gram SRILM | Armsoftech.air | Official leaderboard | |
| 8 | Scribetech | 7.76% | Open Task, Kaldi Chain - Pretrained AM + LM(Ext data) | Scribetech | Official leaderboard | |
| 9 | CDAC-Pune | 8.35% | Open Task, Kaldi Chain Model - Fine tuned AM/LM with train/dev | CDAC-Pune | Official leaderboard | |
| 10 | IITG | 10.3% | Closed Task, MFCC + Pitch, Kaldi DNN-HMM AM + 4gram LM | IITG | Official leaderboard | |
| 11 | IRL-Bangalore | 10.3% | Open Task, E2E Model | IRL-Bangalore | Official leaderboard | |
| 12 | Zen3 Info Solutions | 11.9% | Closed Task, Kaldi TDNN E2E + SentencePiece Lexicon | Zen3 Info Solutions | Official leaderboard | |
| 13 | NIT-Kurukshetra | 37.8% | Closed Task, ESPnet Hybrid CTC/attention E2E (BLSTM) ASR | NIT-Kurukshetra | Official leaderboard | |
| 14 | IISc | 42.2% | Closed Task, E2E: wav2letter | IISc | Official leaderboard |
No lab model card reports it yet
We read the headline results of 540 launch posts from 57 labs, and none of them list this benchmark.
More in Audio and video
- Video-MMMU In use 300 expert-level lecture videos with questions testing perception, comprehension and adaptation of knowledge. Accuracy.
Qwen 3.8 Max 88.7%
- LVBench In use Hours-long video understanding with multiple-choice questions on reasoning, grounding and summarization.
Gemini 4 Argon 91.7%
- MMVU In use 3,000 expert-annotated questions on videos across 27 disciplines. Multiple-choice and open-ended accuracy.
Qwen 3.8 Max 82.4%
- MLVU In use Multi-task long video understanding (minutes to hours) with multiple-choice and generation tasks. Accuracy.
Qwen 3.8 Max 90.8%
- FLEURS In use Models transcribe read speech in 102 languages built from FLoRes-101 sentences. Model cards usually report word error rate, so lower scores are better.
Phi-4-multimodal 4.00%
- VoiceBench In use Spoken instructions with speaker, environment and content variation. Voice assistant responses scored by GPT judge or rules.
Qwen3.5-Omni-Plus 93.1%
Data audit
All 8 checks pass. Every benchmark
Catalog record 2 pass
- Pass Task access All tasks are public.
- Pass Confidence Checked against a primary source.
Scores 2 pass 3 notes
- Pass Independent runs Scores from the official leaderboard (14 entries).
- Pass Models scored 14 entries on its leaderboard, none a model we track.
- Note Newest score No scores yet.
- Note Frontier models No models scored yet.
- Note Headroom Lower is better, so there’s no ceiling to reach.
Leaderboard import 3 pass
- Pass Importer We read the board on sites.google.com: 14 entries at the last read, none a model we track.
- Pass Last read November 20th, 2026.
- Pass Read worked The last read worked.
Consistency 1 pass 1 note
- Note Lab claims No model has both a lab number and an independent run.
- Pass Known issues No published audit found problems.