Understand this story
Chariot released the real-time speech recognition model Chariot-1-ASR Realtime on October 22, supporting 10 Indian languages including Hindi, Marathi, Bengali, and Tamil as well as Indian English, and it is now live on the Chariot platform. According to Analytics India Magazine, the company says the model was tested on 79 public test sets and about 153,000 recordings; in batch mode its error rate is 26% lower than Google Gemini 3.5 Transcribe and 9% lower than Sarvam Saaras v4; in streaming mode it is 17% lower than Sarvam Saaras v4 and 57% lower than ElevenLabs Scribe v2 Realtime, and these results all come from Chariot's own testing. The model offers balanced and high-accuracy modes, can automatically detect language, and supports writing borrowed English words in Latin letters or in the corresponding language's script. There are currently no subsequent development facts.
AI-generated from public reports
Latest development · Report
Chariot releases Chariot-1-ASR Realtime, supporting real-time speech recognition in 10 Indian languagesDevelopments
1 confirmed developments
Chariot releases real-time speech recognition model Chariot-1-ASR Realtime
Source:Analytics India Magazine
Reports and discussion timeline
Browse reports and discussion by time. Their evidence levels remain distinct.
Oct 23
Analytics India Magazine
Chariot releases Chariot-1-ASR Realtime, supporting real-time speech recognition in 10 Indian languages
Chariot released the real-time speech recognition model Chariot-1-ASR Realtime on October 22, supporting 10 Indian languages including Hindi, Marathi, Bengali, and Tamil, as well as Indian English, and it is now available on the Chariot platform. The company says it tested on 79 public test sets and about 153,000 recordings; in batch mode the error rate is 26% lower than Google Gemini 3.5 Transcribe and 9% lower than Sarvam Saaras v4; in streaming mode it is 17% lower than Sarvam Saaras v4 and 57% lower than ElevenLabs Scribe v2 Realtime, and these results all come from Chariot's self-testing. The model offers balanced and high-accuracy modes, can automatically identify the language, and supports writing borrowed English words in English letters or the corresponding language's script.