AI & RoboticsNews

New AI search engine Upend emerges from stealth, powered by 100 LLMs

Upend: Revolutionizing Search with Large Language Models

As Google and Microsoft continue to revamp their search engines with generative AI models, smaller players are going in all to challenge them with their AI-first offerings.

Case in point: Upend, a Canadian startup that has just emerged from stealth to empower students and professionals with gen AI search driven by some of the best large language models (LLMs) out there.

Upend started out as a summer project by Jeevan Arora from the University of Toronto School. After a positive initial response, he evolved it into a full-fledged platform that enterprise teams can sign up for. It works very much like Perplexity, which many believe currently leads the space when it comes to AI search (with 169 million monthly queries).

“My goal is to make advanced gen AI models more affordable, thereby democratizing access and ensuring everyone can harness the tools of tomorrow,” the CEO noted in a statement.

What makes Upend different?

At the core, Upend offers users a gen AI search bar where they can select any LLM from the options on offer and ask it a question about work or everyday life. The model uses the query and combines it web search or the select source to provide grounded answers, complete with citations to help users go back to the source of the answer.

Upend AI platform

The whole thing works very much like Perplexity, although Upend clearly appears like an early-stage product.

Perplexity, on its part, offers more comprehensive capabilities including AI image search and data retention controls.

However, Upend claims to differentiate itself with a bigger stock of models to choose from. Unlike Perplexity, which has about five mainstream LLMs to choose from, Upend has a package of 100.

This includes all big and small closed and open models, including general-purpose models from OpenAI, Claude and Mistral as well as task-specific ones like Meta’s Code Llama and Deepseek Coder. The new platform also provides an option to base answers on Wikipedia, which is not the case with Perplexity.

“We currently support more office file formats for analysis like Word and Excel. One large differentiator is our approach to pricing – we really want to democratize access by charging based on total team usage and not charging based on the number of users on a team,” Arora told VentureBeat.

Currently, the team plan of the product comes at $20/month. This gives access to all the models in its library up to a certain token threshold. After that, users have to pay based on their usage. There’s also a discounted student plan priced at $5/month.

In contrast, Perplexity’s consumer-centric Pro plan comes at $40/month or $400/year per seat, while the ChatGPT Team plan costs $25/month per user when billed annually, or $30/month per user when billed monthly. Upend claims it can drive up to 90% cost reduction when compared to these offerings.

Handling tasks and more in the pipeline

While the product is very new and many kinks remain to be ironed out, especially on the UX front, Arora envisions Upend can eventually become a full-fledged task engine that would not only provide answers but also execute tasks for users.

“As an example, a user searching for “How to convert my PNG image to JPG” would get an answer, but also get prompted to complete the task for them,” the CEO added.

For now, he plans to build up the model library and add much-needed capabilities into the product, including single-sign-on (SSO) to make it easier for businesses to onboard entire teams. He is exploring launching a mobile of Upend, along with an API.

That said, with offerings like Upend and Perplexity gaining traction, the market for traditional search is expected to take a hit. According to Gartner, traditional search engine volume will drop 25% by 2026, with search marketing losing market share to AI chatbots and other virtual agents.

“Generative AI solutions are becoming substitute answer engines, replacing user queries that previously may have been executed in traditional search engines. This will force companies to rethink their marketing channels strategy as GenAI becomes more embedded across all aspects of the enterprise,” Alan Antin, VP analyst at Gartner, said in February.

Author: Shubham Sharma
Source: Venturebeat
Reviewed By: Editorial Team

Related posts
AI & RoboticsNews

OpenAI expands ChatGPT Canvas to all users

AI & RoboticsNews

OpenAI’s o1 model doesn’t show its thinking, giving open source an advantage

AI & RoboticsNews

Singapore startup Sapient enters global enterprise AI race with new model architectures

DefenseNews

Marines hindered by Navy’s amphibious warfare ship maintenance delays

Sign up for our Newsletter and
stay informed!