SUTRA Models
SUTRA-Online
SUTRA-ONLINE
models are online LLMs pioneered by Two Platforms. They are online LLMs because they can use knowledge from the internet, and thus leverage the most up-to-date information when forming response. By providing SUTRA-Online knowledge from web, they can accurately respond to time sensitive queries, unlocking knowledge beyond its training corpus. This means that SUTRA models can respond to queries like “Who won in yesterday’s India vs England cricket match?” accurately, whereas offline models like Llama2, GPT-3.5 etc. suffer from knowledge cut-off dates. Following example shows how SUTRA-Online series models respond to queries with up-to-date information, without suffering from knowledge cut-off dates from training set.
SUTRA-ONLINE
response for “Who won in yesterday’s India vs England cricket match?” (as of 02/20/2024)
GPT-3.5
response for “Who won in yesterday’s India vs England cricket match?” (as of 02/20/2024)
SUTRA-Light
SUTRA-LIGHT
models are extensively multilingual language models adept at executing instructions in 50+ languages, with special emphasis on Latin (English, Italian, French), Indic Languages (Hindi, Gujarati, Tamil etc.) and far eastern languages (e.g. Korean). Developed using a diverse mix of instructions from our proprietary and leading open-access datasets, it achieves state-of-the-art performance across numerous multilingual benchmarks. Following example of response in Hindi showcases the nuance with with SUTRA-Light
can answer, understanding the context specificity of language.
SUTRA-LIGHT
response for “बहुत बढ़िया! स्मोक्ड सैल्मन क्या है? (Great, what is a smoked salmon?)”
GPT-3.5
response for the same question
SUTRA-Avatar
SUTRA-AVATAR
are visual generative AI models that enable realtime interactions with photorealistic AI avatars. These models can create photorealistic AI characters in full HD resolution, complete with emotions, gestures, and synchronized natural-sounding voice. SUTRA-AVATAR
allows extensive customization, offering a range of appearances, expressions, and emotions to facilitate authentic interactions with synthesized photorealistic artificial characters.
SUTRA-AVATAR
can be driven by either text and/or audio and provides the ability to perform realtime and live video calls with AI characters. Models can be accessed either in the form of videos or as a live video stream. The models are light weight and can be accessed via API endpoint. Specific instances of SUTRA-AVATAR
are ultra light weight and can run locally on-device.