In this page, we showcase some of the key characteristics of SUTRA models and how they differ from GPT-3.5 and GPT-4. Both GPT-3.5 and GPT-4 are impressive in terms of their capabilities to answer questions, summarize text and number of other language generation capabilities. Although impressive in their capabilities, they still hallucinate when they have to respond with queries that require real-time information.

Question: ”How was the state of the union address?” [asked on 03/08/2024 8AM PST, the morning after SOTU address was delivered by President Joe Biden]

Response from SUTRA-ONLINE: It understands the question and gets most recent information from internet and formulates and up-to-date response with minimal latency.

President Joe Biden delivered 2024 state of the union
address yesterday. He talked about issues on health care,
climate change, humanitarian support for Gaza and
War in Ukraine and updating tax rates for rich.
[latency: 500 ms]

Response from GPT-3.5: The model suffers from knowledge cutoff and is not able to formulate up-to-date response. Furthermore, it takes about 4X more time.

I'm sorry, but I cannot provide real-time information
or updates as my training only includes data up to January
2022. Please consider checking current news.
[latency: 1800 ms]

Response from GPT-4: The model suffers from knowledge cutoff and hallucinates about 2023 SOTU address while taking about 10X more time

In his 2023 State of the Union address delivered on
February 7, President Joe Biden emphasized themes of
progress, resilience, and bipartisanship and recovery from
COVID-19
[latency: 5100 ms]
SpeedSUTRA vs Mistral Models