| command | Operation: chat, generate-image, generate-audio, analyze-image, rent-gpu, list-gpus, terminate-gpu | yes | — |
| api-key | Hyperbolic API key | yes | — |
| api-url | API base URL override | no | https://api.hyperbolic.xyz |
| model | Model ID (e.g. "deepseek-ai/DeepSeek-V3", "meta-llama/Llama-3.3-70B-Instruct") | no | — |
| messages | Chat messages as JSON array (e.g. [{"role":"user","content":"Hello"}]) | no | — |
| temperature | Sampling temperature (0.0-2.0) | no | — |
| top-p | Nucleus sampling threshold (0.0-1.0) | no | — |
| max-tokens | Maximum tokens to generate | no | — |
| response-format | Response format as JSON (e.g. {"type":"json_schema","json_schema":{...}}) | no | — |
| seed | Random seed for reproducibility | no | — |
| stop | Comma-separated stop sequences | no | — |
| tools | Tool definitions as JSON array for function calling | no | — |
| prompt | Text prompt for image generation or question for vision analysis | no | — |
| height | Image height in pixels (default: 1024) | no | — |
| width | Image width in pixels (default: 1024) | no | — |
| steps | Number of diffusion steps (default: 25) | no | — |
| lora | LoRA style adapter name (e.g. "Pixel_Art", "Logo", "Sci-fi") | no | — |
| lora-weight | LoRA influence weight (0.0-1.0, default: 0.8) | no | — |
| text | Text to convert to speech | no | — |
| language | TTS language: EN, ES, FR, ZH, JP, KR (default: EN) | no | — |
| speaker | TTS speaker/accent (e.g. "EN-US", "EN-BR", "EN-AU") | no | — |
| speed | TTS speed multiplier (0.1-5.0, default: 1.0) | no | — |
| image-url | URL of image to analyze | no | — |
| image-base64 | Base64-encoded image to analyze | no | — |
| gpu-type | GPU type for rental (e.g. "H100-SXM", "A100-SXM", "RTX-4090") | no | — |
| gpu-count | Number of GPUs (1, 2, 4, or 8) | no | 1 |
| instance-id | Instance ID for terminate-gpu | no | — |