Korean AI That Surpassed Google Emerges… 'Scinapse AI' to Launch in Q3

Pluto Labs Unveils 'Scinapse AI'

Development Costs, Only 1/10th of Existing AI

Outperformed Google and Anthropic in Performance Tests

Launch in Q3… Targeting the Global Market

South Korean artificial intelligence (AI) startup Pluto Labs has succeeded in developing a 'low-cost, high-efficiency' AI model. With better performance than U.S. big tech’s AI models in blind tests, there is growing anticipation that it could become the 'Korean version of DeepSeek.'

On the 11th, Pluto Labs unveiled its own AI model, 'Scinapse AI.' Scinapse AI is an 'AI scientist model' that supports automating scientific research and is scheduled for official launch in the third quarter. Pluto Labs stated, “Pilot programs of Scinapse AI are being tested at major universities in the U.S., Europe, and Asia,” adding, “After the official launch, we plan to actively target the global research market, including universities and institutes.”

The core of Scinapse AI is considered its 'efficiency.' Specifically, its development cost was only 1/10th that of top-tier AI models from Google and U.S. startup Anthropic. This is thanks to adopting a strategy of 'strategic delegation,' which establishes an environment where the AI can operate most efficiently before it begins reasoning. Strategic delegation is a development approach that focuses the AI solely on creative reasoning and idea generation grounded in essential information.

For processing and filtering large volumes of data—which would otherwise be inefficiently handled—Pluto Labs’ academic search service, 'Scinapse,' takes over. This contrasts with the training method of Google’s AI scientist model 'Co-scientist (Co-scientist),' which feeds vast amounts of data into the AI model during reasoning.

An industry insider explained, “The existing AI training method incurs tens of dollars in costs for just a single inference, which limits practical applications in real research,” adding, “(The strategic delegation method) is akin to a CEO not directly handling every task, but instead specialized teams in each field maximizing their strengths for greater efficiency.”

Dominance in Blind Performance Evaluations

Scinapse AI outperformed in blind performance evaluations where global AI models such as OpenAI’s O3, Google’s Gemini 2.5 Pro, and Anthropic’s Claude Opus 4 served as judges. Notably, in assessments of 61 research topics, it ranked first among all AI models in the categories of ‘scientific validity’ and ‘verifiability.’

A Pluto Labs official commented, “What’s notable in the blind test is not simply that Scinapse produced creative or interesting ideas, but that it formulated hypotheses that could actually be experimentally tested and implemented within three to five years,” adding, “This means Scinapse AI has moved beyond generating ‘plausible fiction’ like previous AIs and entered the domain of ‘practical scientific research.’”

Scinapse AI also adopted a 'multi-verification system' to tackle the issue of 'AI hallucinations'—phenomena where AI presents false information as fact. Scinapse AI generates ideas based on a database of over 260 million papers, and the generated ideas are then cross-checked against existing studies for overlaps using multiple keyword searches.

“Surpassed Google… A Symbolic Event for Technological Independence”

The academic community's response to Scinapse AI has also been positive. Professor Cho Chang-shin of the Department of Chemical Engineering at POSTECH (Pohang University of Science and Technology) stated, “Scinapse AI completes prior literature reviews—normally taking human researchers several weeks—in just a few minutes,” adding, “its accuracy is actually even higher.” He further remarked, “This is an innovation that fundamentally changes the starting point of research planning, going beyond an AI that just proposes ideas.”

Industry observers view the achievement as especially meaningful in the context of a 'David vs. Goliath' showdown. Pluto Labs, established in 2017, has raised a cumulative investment totaling ₩5 billion. Considering that the total investment raised by U.S. big tech company Anthropic reaches $14.3 billion (approximately ₩19.7 trillion), this highlights the extreme efficiency of Pluto Labs’ AI development.

Of particular note: Pluto Labs already owns 'Scinapse,' a proven platform used by 170,000 researchers worldwide. This offers a powerful foundation for continuous improvement through real user feedback, beyond just technological development.

Seojoon Kim, CEO of global Web3 venture capital firm Hashed, noted, “The fact that a small Korean startup beat Google—the symbol of immense capital—based on objective benchmarks, is a symbolic event of technological independence,” adding, “Scinapse AI marks a milestone for Korea’s AI ecosystem, proving that creative problem-solving, rather than limitless capital, can win in the global technology race.”