The latest service changes have not yet been reflected in this content. We will update the content as soon as possible. Please refer to the Korean version for information on the latest updates.
Available in Classic and VPC
HyperCLOVA X, provided by CLOVA Studio, is NAVER's generative AI that best understands Korean culture and context.
HyperCLOVA X models
The HyperCLOVA X model lineup includes:
- HCX-007: A hybrid reasoning model capable of generating responses that reflect deep reasoning processes. It can be used for solving complex tasks such as mathematics, science, language reasoning, and writing.
- HCX-005: A multimodal model capable of interpreting and understanding images, with strong performance in following a wide range of instructions.
- HCX-DASH-002: A lightweight model optimized for high speed and throughput, suitable for relatively simple tasks.
| Item | HCX-007 | HCX-005 | HCX-DASH-002 | HCX-003 | HCX-DASH-001 |
|---|---|---|---|---|---|
| Supported input types | Text | Text/Image | Text | Text | Text |
| Context length | 128,000 | 128,000 | 32,000 | 8,192 | 4,096 |
| Max input tokens | 128,000 | 128,000 | 32,000 | 7,600 | 3,500 |
| Max output tokens | 32,768 | 4,096 | 4,096 | 4,096 | 4,096 |
| API | |||||
| Tuning support | ✕ | ○ | ○ | ○ | ○ |
| Function calling support | ○ | ○ | ○ | ✕ | ✕ |
| Structured Outputs support | ○ | ✕ | ✕ | ✕ | ✕ |
| Reasoning (Thinking) support | ○ (Hybrid) | ✕ | ✕ | ✕ | ✕ |
| Amount | View pricing | ||||
| Usage limits (QPM/TPM) | Usage control policy | ||||
Note
- Context length refers to the sum of input and output tokens the model can process.
- Image input, tuning, function calling, structured outputs, and reasoning cannot be used simultaneously. Specify the relevant parameters when making a request. For details, refer to the API Guide.