From tech giants to startups, South Korean gamers are growing extensive language fashions adapted to their very own language and tradition, in a position to compete with international heavyweights like OpenAI and Google.
Final month, the country introduced its maximum formidable sovereign AI initiative to this point, pledging ₩530 billion, (about $390 million), to 5 native firms development large-scale foundational fashions.
The transfer underscores Seoul’s want to chop reliance on overseas AI applied sciences, hoping to enhance nationwide safety and stay a tighter keep watch over over information within the AI technology.
The organizations picked through the Ministry of Science and ICT to compete had been LG AI Analysis, SK Telecom, Naver Cloud, NC AI, and the startup Upstage.
Each six months, the federal government will overview the primary cohort’s development, lower underperformers, and proceed investment the frontrunners till simply two stay to guide the rustic’s sovereign AI force.
Each and every participant is bringing a unique merit to South Korea’s AI race. TechCrunch spoke with a number of of the chosen firms about how they plan to tackle OpenAI, Google, Anthropic and the remainder on their house turf. NC AI declined to remark.
LG AI Analysis: Exaone
LG AI Analysis, the R&D unit of South Korean massive LG Workforce, gives Exaone 4.0, a hybrid reasoning AI fashion. The newest model blends large language processing with the complicated reasoning options first offered within the corporate’s previous Exaone Deep fashion.
Techcrunch match
San Francisco
|
October 27-29, 2025
Exaone 4.0 (32B) already ratings somewhat smartly in opposition to competition on Synthetic Research’s Intelligence Index benchmark (as does Upstage’s Sun Pro2). Nevertheless it plans to reinforce and transfer up the ranks via its deep get right of entry to to real-world {industry} information starting from biotech to complicated fabrics and production.
It’s coupling that information with a focal point on refining the information ahead of feeding to the fashions to coach. As a substitute of chasing sheer scale, LG desires to make all the procedure extra clever, so its AI can ship genuine, sensible worth that is going past what general-purpose fashions can be offering. “That is our elementary means,” co-head Honglak Lee informed TechCrunch.
LG is bettering its fashions by way of acquainted ways: providing them via APIs, then the use of the real-world information generated through customers of the ones services and products to coach the fashion to reinforce.
“As LG’s fashions reinforce, our companions can ship higher services and products, which in flip generate larger financial worth or even richer information,” he mentioned.
On the other hand, as a substitute of chasing large GPU clusters, LG AI Analysis is specializing in potency, getting probably the most out of each chip, and developing industry-specific fashions, he discussed. The function isn’t to outspend the worldwide giants however to outsmart them with high-performing, but extra environment friendly, AI.
South Korea’s telco massive SK Telecom (SKT) introduced its non-public AI agent A. (pronounced A-dot) provider long ago in past due 2023 and simply rolled out its new extensive language fashion, A.X, this July.
Constructed on most sensible of the Chinese language open supply fashion from Alibaba Cloud, Qwen 2.5, A.X 4.0 is available in two fashions, a hefty 72-billion-parameter model and a lighter 7B model.
SK says that A.X 4.0 processes Korean inputs about 33% extra successfully than GPT-4o did, underscoring its native language edge. (OpenAI’s GPT 5.0 comparability information isn’t to be had.) SKT additionally open sourced its A.X 3.1 fashions previous this summer season. In the meantime, the A. provider gives options like AI name summaries and auto-generated notes. As of August 2025, it’s already pulled in about 10 million subscribers.
SK’s edge is its versatility, as it has get right of entry to to data from its telecom community starting from navigation to taxi-hailing.
“SK Telecom’s position is to behave as a bridge between state-of-the-art fashion analysis and real-world have an effect on. With our telecom infrastructure, in depth consumer base and confirmed provider like A., we deliver AI at once into on a regular basis existence, whether or not in customer support, mobility, or production,” Taeyoon Kim, head of the root fashion place of business at SK Telecom, informed TechCrunch.
SK Telecom could also be making an investment in AI infrastructure, the use of GPUaaS, South Korea’s greatest GPU-based provider, and development a brand new hyperscale AI information middle with AWS. No matter it lacks, it’s partnering to procure.
“We’re development a full-stack ecosystem with Korean AI chipmaker Rebellions, securing depended on information partnerships via paintings with the federal government and universities, and fostering an international analysis community,” mentioned Kim. “That comes with tasks like our collaboration with MIT (MGAIC), which applies basis fashions to complicated production and battery and semiconductor innovation.”
Naver Cloud: HyperCLOVA X
Naver Cloud, the cloud services and products arm of South Korea’s main web corporate, offered its extensive language fashion, HyperClova, in 2021. Two years later, it unveiled an upgraded model, HyperCLOVA X, in conjunction with new merchandise powered through the era: CLOVA X, an AI chatbot, and Cue, a generative AI-driven seek engine located as a rival to Microsoft’s CoPilot-enhanced Bing and Google’s AI Evaluate. It additionally unveiled this 12 months its multimodal reasoning AI fashion, HyperCLOVE X Suppose.
Naver Cloud believes the actual energy of LLMs is to function “connectors” linking legacy programs and siloed services and products to reinforce usefulness, in step with a Naver spokesperson.
Naver sticks out as Korea’s handiest corporate — and some of the few on this planet — that may really declare to have an “AI complete stack.” It constructed its HyperCLOVA X fashion from scratch and runs the huge information facilities, cloud services and products, AI platforms, programs, and client services and products that deliver the era to existence, the spokesperson defined.
Very similar to Google — however tuned for South Korea — Naver is embedding its AI into core services and products like seek, buying groceries, maps, and finance. Its merit is real-world information. It’s AI Buying groceries Information, as an example, gives suggestions in keeping with what other folks if truth be told wish to purchase. Different services and products come with CLOVA Studio, which we could companies construct customized generative AI, and CLOVA Carecall, an AI-powered check-in provider geared for seniors dwelling on my own.
The Naver spokesperson says besting international AI giants like OpenAI and Google hinges on two issues: perfecting its “recipe” for fashions and securing the capital to scale them. Even so, reasonably than chasing measurement, the corporate emphasizes sophistication, arguing its AI is already globally aggressive at similar scales.
Upstage’s Sun Professional 2
Upstage is the one startup competing within the venture. Its Sun Professional 2 fashion, introduced ultimate July, used to be the primary Korean fashion known as a frontier fashion through Synthetic Research, placing it within the ring with OpenAI, Google, Meta, and Anthropic, in step with Quickly-il Kwon, govt vice chairman at Upstage.
Whilst maximum frontier fashions have 100 billion to 200 billion parameters, Sun Professional 2 — with simply 31 billion — plays higher for South Koreans and is more cost effective, Kwon informed TechCrunch.
“Sun Professional 2 has outperformed international fashions on main Korean benchmarks. With this venture, Upstage objectives to succeed in a Korean language efficiency of 105% of the worldwide same old,” Kwon mentioned.
Upstage objectives to tell apart itself through specializing in genuine trade have an effect on, now not simply benchmarks, he mentioned. So it’s growing specialised fashions for industries like finance, legislation, and drugs, whilst pushing to construct a Korean AI ecosystem led through “AI-native” startups.
