Open ASR Leaderboard Adds Hindi and Indian English, With 4,888 Speakers

Hugging Face and Voice Arena announced on August 28, 2026 that they have added two evaluation sets for Indian English and Hindi, Monsoon en-IN and Monsoon hi-IN, to the Open ASR Leaderboard¹.

• Hindi is the first Indic language on the leaderboard’s multilingual tab, which until now covered only European languages¹
• Each language ships a public split for self-scoring and a private split withheld to limit benchmark-specific optimisation. The four splits are speaker-disjoint and comprise 4,888 speakers, with 12 attributes recorded per speaker¹
• Sizes are 5.62 hours (1,444 speakers, 428 districts) and 5.58 hours (1,405 speakers, 420 districts) for Indian English, and 1.33 hours (468 speakers, 202 districts) and 4.47 hours (1,571 speakers, 295 districts) for Hindi¹
• The Hindi sets use a lattice of accepted spellings as the reference rather than a single correct string¹

The stated design goal is to vary along nine axes: geography, age, gender, vocabulary, devices, acoustic environments, speech type, speech rate, and the existence of multiple valid transcripts for the same audio¹. The post cites prior work showing that ASR error rates are not evenly distributed across the people using these systems, and notes that conventional test sets record what was said while retaining almost nothing about who said it¹. If you deploy speech recognition outside English-speaking markets, that is a useful reminder that a single leaderboard score may not describe your own user base.

Sources

1. The Open ASR Leaderboard Adds Its First Global South Language - Hugging Face official blog (August 28, 2026)