EXAONE Omni Inspect checks factory quality without retraining
LG put its industrial AI stack on stage in Seoul this morning, and the interesting part isn't the model names — it's that the inspection model no longer needs to be retrained when the product changes. Meanwhile Chinese banks started underwriting AI companies on the strength of how many tokens they burn.
LG AI Research unveiled "expert AI" for manufacturing, science and finance at its AI Talk Concert 2026 today at LG Science Park in Seoul. The headline release is EXAONE Omni Inspect, an autonomous quality-inspection model that runs high-accuracy checks without the retraining cycle that earlier inspection systems required every time a production process or a product's exterior changed; it sits alongside EXAONE Tabular, which forecasts process conditions and quality data. In science, LG laid out a self-driving chemistry laboratory it is building with LG Chem, where robots and AI design compounds, predict synthesis outcomes and run the experiments. Finance runs through EXAONE Business Intelligence, live since earlier this year with London Stock Exchange Group and Koscom partnerships, now folded into EXAONE Forecast and pointed at bonds, commodities, and customers in North America, Europe and the Middle East. Co-head Lim Woo-hyung framed the goal as autonomous factories "where an entire plant operates as a single intelligence," on top of a robot foundation model — the "brain" for robots acting in the physical world. Six years and 1,080 patent applications later, this is what a sovereign-model lab looks like when it stops competing on chatbot benchmarks and sells to its own factories instead.
Beijing's Economic-Technological Development Area booked its first batch of "token loans" at close to 2 billion yuan ($280 million), reported by The Beijing News and confirmed by the municipal government. Six banks — Agricultural Bank of China, China CITIC, Industrial Bank, Bank of Beijing, Minsheng and China Huaxia — extended the credit to AI supply-chain firms in Yizhuang. The underwriting metric is token consumption plus compute-service contract value and receivables, replacing collateral and profit statements that asset-light AI startups don't have. Xinhua first flagged the model in late August as a broader shift in bank risk assessment from "looking at assets" to "looking at data"; Chongqing is now consulting on "compute banks" and token vouchers, and Chengdu proposed a 100 million yuan annual token-voucher pool with up to 2 million yuan per firm. The read: whoever controls token-metering infrastructure becomes the credit bureau for the AI economy, and China's local governments are moving faster on that than any western regulator is moving on model safety.
What to watch: whether Microsoft ships the first-party MAI models it promised for today, and whether LG's robot foundation model shows up as a product rather than a roadmap slide.
Should a bank lend against a company's token consumption — is that a real asset or a burn-rate trophy? Tell us in the comments.
