FEBRUARYASIA BUSINESS OUTLOOK8NEWSROOMLG AI RESEARCH PATENTS EXAONE DISCOVERY AI FOR SCIENCELG AI Research, a division of South Korea's LG Group, has obtained a patent for its agentic AI platform, EXAONE Discovery, which is designed to accelerate scientific research by processing complex data significantly faster than existing systems.The patent covers core technologies including Optical Chemical Structure Recognition (OCSR), enabling the AI to read text, formulas, tables, and images directly from scientific papers and patent filings.LG claims the platform reads chemical structures over 100× better than current models, backing its performance advantage.EXAONE Discovery is already being deployed across multiple industries--supporting product development in cosmetics, batteries, and pharmaceuticals.Key Highlights· LG patents EXAONE Discovery, an AI platform for scientific data analysis· EXAONE applied in cosmetics, batteries, and pharmaceuticals with advanced OCSR· LG has filed 573 AI-related patents and is expanding platform deploymentNotable use cases include identifying new solvent candidates for electrolytes in next-generation lithium-sulfur (Li-S) batteries and predicting neoantigens to advance personalized cancer vaccines.Another LG Group unit, LG Innotek, uses the platform to optimize camera lens and sensor processes, reducing optimization time by over 50 percent.Since 2022, LG AI Research reports having filed 573 AI-related patent applications to protect its growing portfolio of AI technologies. Beyond EXAONE, the group has discussed an AI framework behind what it describes as the world's first fully AI-managed ETF listed on the NYSE.To broaden EXAONE's adoption, LG is collaborating with companies like Nota AI to incorporate compression and optimization tools, and Letser to support secure corporate AI verification.Internal process wins, patent filings, and strategic partnerships illustrate how large industrial players can transform in-house AI R&D into broader commercial opportunities and revenue streams. CHINESE AI MODELS POWER OPENCLAW's LOW-COST AGENT PUSHChinese open-source AI models such as Moonshot AI's Kimi and MiniMax are being integrated into OpenClaw, a rapidly growing open-source AI agent, as users prioritize cost efficiency and performance.OpenClaw announced on January 30 that it now offers Kimi K2.5 and Kimi Coding for free, while also supporting models from MiniMax, a Chinese foundation model developer.Analysts highlight that Chinese models are gaining traction due to affordability. Kimi reportedly costs around US$0.6 per million input tokens, far cheaper than U.S. premium models like Anthropic Claude Opus 4.5, priced at US$5 per million tokens.This pricing gap is significant for users running autonomous AI agents, where token usage can surge unexpectedly.Key Highlights· OpenClaw integrates Moonshot Kimi and MiniMax for cheaper AI runs· Chinese models cost up to 8× less per token than premium U.S. AI· Always-on AI agents can trigger unexpected high API billsOpenClaw, launched in late 2025, is a free, open-source AI agent capable of connecting to online services and executing diverse tasks.It runs locally on Mac, Windows (WSL2), or Linux, though deployments often rely on third-party model APIs, making API calls the main cost driver.Features like persistent memory and a "heartbeat" function enable always-on automation, which can rapidly consume tokens. One developer reportedly faced a $623 monthly API bill, underscoring why lower-cost models matter.Experts say OpenClaw's adoption of Kimi and MiniMax reflects a broader AI pricing war, where high-volume agent workloads pressure expensive model providers. While premium models offer top-tier performance, "value for money" increasingly drives adoption in the AI agent ecosystem.
< Page 7 | Page 9 >