Summary
Naver Clova AI offers a comprehensive AI platform centered around its HyperCLOVA X large language model, optimized for the Korean language and culture. It provides tools for developers and enterprises to build, customize, and deploy AI services, including conversational AI, multimodal capabilities, and fine-tuning options. The platform emphasizes ethical AI development and offers various solutions for different business needs.
Features9/13
See allMust Have
5 of 5
Conversational AI
API Access
Safety & Alignment Framework
Fine-Tuning & Custom Models
Enterprise Solutions
Other
4 of 8
Image Generation
Multimodal AI
Research & Publications
Security & Red Teaming
Code Generation
Synthetic Media Provenance
Threat Intelligence Reporting
Global Affairs & Policy
PricingTiered
See allBasic
- Access to HyperCLOVA X API
- Early access to new features
- Usage and performance not guaranteed
Exclusive
- Everything in Basic plan
- Advanced tuning
- Dedicated infrastructure
- Own GPU and HyperCLOVA X model
- TPM guarantees
Neurocloud for HyperCLOVA X
- Fully managed service
- Hardware and software run at on-premises data center
- Hybrid cloud environment
- Configure on-premises networks
- Customized HyperCLOVA X model
- Model training on business data
- Supervised fine-tuning (SFT)
- Further pretraining (FP)
- Reinforcement learning from human feedback (RLHF)
Rationale
Naver Clova AI, particularly through its HyperCLOVA X and CLOVA Studio offerings, directly aligns with all must-have features. It provides conversational AI (CLOVA X chatbot), API access (HyperCLOVA X as an API, CLOVA Studio for LLMOps), safety and alignment (explicit mention of AI ethics principles and unlearning techniques for safe image generation), fine-tuning and custom models (CLOVA Studio's tuning methods like PEFT, SFT, RLHF), and enterprise solutions (CLOVA Studio Exclusive and Neurocloud for HyperCLOVA X, and custom generative AI platforms for enterprises like Bank of Korea). Additionally, it offers image generation (AI Eraser, image understanding in HyperCLOVA X), multimodal AI (understanding image and audio data), research and publications (over 300 papers published), and security/red teaming (unlearning techniques for safe AI image generation, security compliance in Neurocloud).
