HyperCLOVA X HCX-003
Naver · South Korea
- Params
- 204B
- Context
- 65,536
- Min VRAM
- 200 GB
- GGUF
- No
Exceptional Korean text generation with cultural nuance.
Region slug: koreaAI models, vendors, and hardware buying realities from South Korea. Curated by myaihardware.com.
South Korea is one of the few places on earth where you can buy high-bandwidth memory (HBM) and advanced foundry capacity for AI chips outside of TSMC's Taiwan, a hardware-availability reality that directly de-risks your supply chain. Its AI builders overwhelmingly deploy open-weight models like Llama and Mistral, not proprietary APIs, making the market a strong signal for inference-heavy hardware demand. The underrated angle is that Korea's dense data center buildout is driven by domestic hyperscalers (Naver, Kakao) not just Samsung/LG, creating a unique channel for edge and HPC hardware that ignores the US-centric narrative.
South Korea's sovereign-AI drive is propelled by Naver, LG, and Samsung building domestic foundation models alongside government chip subsidies that aim to reduce reliance on foreign hardware. The evaluation landscape has rapidly matured around native benchmarks like KMMLU, HAERAE, and LogicKor, which capture linguistic and cultural nuance far better than translated English tests. Underlying these efforts, traditional byte-pair encoding stumbles on Hangul's dense syllabic blocks within the CJK range, spurring bespoke Korean-first tokenizers that optimize morphology by treating jamo sequences rather than raw bytes.
Filter by vendor, license, or parameter range.
Naver · South Korea
Exceptional Korean text generation with cultural nuance.
Region slug: koreaNaver · South Korea
Efficient and fast for real-time applications.
Region slug: koreaLG AI Research · South Korea
Excellent bilingual performance with efficient size.
Region slug: koreaLG AI Research · South Korea
Very lightweight and fast, good for on-device.
Region slug: koreaLG AI Research · South Korea
Strong reasoning improvements over 3.0.
Region slug: koreaLG AI Research · South Korea
Top-tier performance among open Korean-English models.
Region slug: koreaLG AI Research · South Korea
Efficient inference via MoE, excels at reasoning.
Region slug: koreaUpstage · South Korea
Packs 13B-level performance in a smaller size.
Region slug: koreaUpstage · South Korea
Extremely fast and low-resource.
Region slug: koreaUpstage · South Korea
Best performance among SOLAR models, good Korean.
Region slug: koreaKT Corporation · South Korea
Deep Korean linguistic knowledge from KT's data.
Region slug: koreaSK Telecom · South Korea
Optimized for Korean conversational AI.
Region slug: koreaKakao · South Korea
Strong Korean generation from Kakao's data ecosystem.
Region slug: koreaEleutherAI · South Korea
Open and reproducible Korean pretrained model.
Region slug: koreaNCSoft · South Korea
Tailored for game narratives and interactive fiction.
Region slug: korea42dot · South Korea
Optimized for automotive and navigation tasks.
Region slug: koreallama.cpp, MLPerf, and inference leaderboards.
Will this model fit on your GPU?
Local LLM setup, homelab, mini-PC builds.
How we benchmark, verify ASINs, and source data.
New to local AI? Start here.
Sourcing & commerce disclosure
Model metadata on this page is compiled from each vendor's public HuggingFace repo, model card, or research paper at time of indexing. Performance figures (KMMLU, MMLU, HAERAE, AraBench, IndicGEM, etc.) link directly to source benchmarks where available; missing or partial values are shown as ", ".
Outbound hardware links (GPUs, mini-PCs, accelerators) are Amazon Associate links using store ID fredoline-20. As an Amazon Associate we earn from qualifying purchases. Prices and availability shown elsewhere on the site are snapshots, not live quotes — verify on Amazon before purchasing. See About / disclosures for the full policy.