Korean startup claims its 'AI scientist' outperforms Gemini 2.5 Pro and Claude Opus 4

In the race to build a true 'AI scientist', Korean startup has announced a new model that it claims achieves superior performance at just one-tenth the computational cost of its competitors. Pluto Labs' Scinapse AI challenges the notion that success in advanced AI requires massive, brute-force computing power.

Pluto Labs says its “Strategic Efficiency” architecture is the key to its success — Scinapse AI delegates data-intensive tasks like literature reviews to its existing academic search platform. This design allows the core AI to focus exclusively on creative reasoning and the generation of novel, testable ideas.

Based on a self-reported blind test, the company boasts that its AI consistently ranked first in “Plausibility” and “Testability” against top models from Google and Anthropic across 61 different scientific topics. Pluto Labs claims this proves its AI delivers “actionable scientific research” rather than “plausible fiction”. To combat AI hallucination, the system is said to be strictly grounded to a database of 260 million academic papers. The blind tests were judged by 3 competing AI models: Gemini 2.5 Pro, Clause Opus 4, and OpenAI o3.

The fact that a small Korean startup has proven superior to Google in an objective benchmark is more than just a technical achievement — it's a symbolic event for innovative leadership. — Simon Kim, CEO of Hashed.

Pluto Labs plans a global launch for Scinapse AI this quarter. The tech community will be watching closely for independent benchmarks to validate the company's bold claims.