Genspark, an artificial intelligence search startup that competes with Google LLC, has reportedly raised $100 million in funding.

Reuters today cited sources as saying that the round was led by a group of U.S. and Singapore-based investors. The deal reportedly values Genspark at $530 million, more than twice what it was worth in June.

Genspark, officially Mainfunc Inc., launched last year with $60 million in initial funding. Its namesake search engine topped 1 million monthly users in November. Sources told Reuters that Genspark now has 2 million users, which was likely one of the factors behind the steep increase in the company’s valuation. 

Genspark’s search engine uses large language models to process queries. When a user types in a question, the service doesn’t return a list of relevant webpages but rather displays a natural language answer. Genspark organizes each of its prompt responses into a webpage called Sparkpages that allows users to ask followup questions.

The company also offers a number of specialized search features. One capability is geared towards browsing e-commerce websites’ product listings. Another tool, Genspark Finance, visualizes data from earnings reports in graphs to ease analysis. 

One of the latest additions to the search engine’s feature set is a capability called Deep Research. It provides more detailed answers to user queries at the cost of increased wait times: Genspark says that a prompt response can take up to 30 minutes to generate. According to the company, the feature spends that time analyzing more than 1 million words’ worth of information from more than a thousand sources.

Under the hood, Genspark’s search engine is powered by what it describes as a mixture-of-agent architecture. The software sends each user query to LLMs from OpenAI, Anthropic PBC and Google. Genspark removes inconsistencies between the models’ responses and then combines them into a single answer.

Today’s report didn’t specify how the company plans to spend its newly raised funding. One possibility is that Genspark could add reasoning-optimized LLMs to the lineup of models it uses to process prompts.

According to a Feb. 10 blog post, the company’s search engine relies on GPT-4o, Claude 3.5 Sonnet and Gemini 2.0 Flash to answer queries. Those are midrange LLMs that balance output quality with hardware efficiency. Adding a more hardware-intensive, reasoning-optimized model such as OpenAI’s o1 to the list could potentially help Genspark enhance its search engine’s output. 

The company’s funding round comes a few weeks after Perplexity AI Inc., a competing AI search startup, closed a $500 million investment of its own. The deal reportedly valued the company at $9 million. Last week, Perplexity launched a feature similar to Genspark’s Deep Research that can generate multipage prompt responses in response to user queries. 

Image: Genspark