Reka Speech is an on-device multilingual speech-to-text and speech-to-speech transcription and translation model.
[REKA SPEECH]
Meaning Made Intelligible








































































































Products
company
resources
[PERFORMANCE]
State-of-the-art performance
Outperforms popular open source models such as Whisper-L v3 which is three times larger on transcription and translation quality.
Translation accuracy (Higher is better)
Translation accuracy (Higher is better)




Transcription error rate (Lower is better)
Transcription error rate (Lower is better)





Resources
Company

Resources
Company

Resources
Company

Resources
Company

Reka Speech is an intelligence layer that transforms multilingual audio data into insights and actions

Reka Vision is an intelligence layer that transforms multimodal data into insights and actions

Reka Vision is an intelligence layer that transforms multimodal data into insights and actions

Reka Vision is an intelligence layer that transforms multimodal data into insights and actions
[SPECIFICATIONS]
Ultra compact and versatile,
for embedding speech AI into
the smallest devices
Ultra compact and versatile, for embedding speech AI into the smallest devices
Ultra compact and versatile, for embedding speech AI into the smallest devices
Run on cellphones, cars, robots, and smart appliances.
This model can be customized further. Talk to Sales ↗
This model can be customized further. Talk to Sales ↗
Less than 1GB memory size
~900 million parameters
Also available via Reka API
English, Japanese, Korean, Chinese, French, Spanish, German, Italian, Portuguese, Arabic
Less than 1GB memory size
~900 million parameters
Also available via Reka API
English, Japanese, Korean, Chinese, French, Spanish, German, Italian, Portuguese, Arabic
Less than 1GB memory size
~900 million parameters
English, Japanese, Korean, Chinese, French, Spanish, German, Italian, Portuguese, Arabic
Also available via Reka API
Less than 1GB memory size
~900 million parameters
Also available via Reka API
English, Japanese, Korean, Chinese, French, Spanish, German, Italian, Portuguese, Arabic