Release Preview
Grok Imagine Image 2.0 is now available as an explicit quality/value option, cheap mode can route budget-sensitive requests automatically, and the Images page shows which model created each asset.
Quick Actions
Jump directly into the workflows covered in this update.
Explicit Grok model selection
POST /content/generate { "project":"campaign", "prompt":"business-team-meeting", "model":"IMAGE_GEN_GROK_IMAGINE_2" }
Budget-aware routing
POST /content/generate { "project":"campaign", "prompt":"product-launch-visual", "mode":"cheap" }
URL model override
https://img.inliner.ai/campaign/business-team-meeting_1200x800.jpg?model=IMAGE_GEN_GROK_IMAGINE_2
Grok Imagine 2 Is Available Across Inliner
- Select Grok Imagine 2 in Studio when you want a strong balance of human realism, prompt adherence, and provider cost.
- Use the exact model code through the API, MCP generate_image tool, or URL model parameter.
- Reference-aware requests can include up to three source images through the OpenRouter-backed adapter.
Cheap Mode Handles Budget-Aware Routing
- General image requests favor Z-Image Turbo.
- Text- and design-heavy requests favor Qwen Image.
- Reference-aware requests use Gemini Flash Lite, while an explicit model selection always takes priority.
The Images Page Now Shows the Model Used
- Each tracked image card displays a concise model badge.
- The image-list API returns both the Inliner model code and provider model identifier.
- Historical images use recorded generation requests as a fallback when model metadata was not saved directly on the asset.
Measured and Auditable Usage
In our production 1024×1024 low-quality test, OpenRouter reported a $0.04 provider cost for Grok Imagine 2. Provider-reported and estimated costs are stored with the generation record so routing decisions can be reviewed over time.