Model Overview
Compare available models and choose the right one for your use case.
| Feature | V3 Pro | V3 Flash | V3.5 Series |
|---|---|---|---|
| Pipeline | 3-stage (LLM + Diff + Renderer) | 2-stage (LLM + Diff) | Enhanced 3-stage |
| Avg Latency | ~2.1s | ~0.4s | ~1.8s |
| Anchor Optimization | Full (Differentiable Renderer) | Partial | Adaptive (Context-Aware) |
| Hex Color Fidelity | Strict | Strict | Strict + Brand Enforcement |
| Layer Grouping | Correct semantic grouping | Correct grouping | Component-aware grouping |
| Alpha Edges | Clean | Clean | Clean |
| Partial Redraw | Supported | Not supported | Multi-layer supported |
| Max Output Size | 512KB | 48KB | 2MB |
| Fine-Tuning | Not available | Not available | Custom fine-tuning |
| Availability | Pro plan and above | All plans | Enterprise only |
When to Use Which Model
V3 Pro
Best for: High-quality production assets
Use V3 Pro when you need the highest structural quality, minimal anchor points, or non-destructive redraw. Ideal for brand assets, icon libraries, and production UI components that will be maintained over time.
V3 Flash
Best for: Real-time and high-frequency generation
Use V3 Flash when latency matters — dynamic UI elements, real-time icon generation, or batch processing of large icon sets. Slightly fewer optimization passes but still produces clean, color-accurate SVG.
V3.5 Series
Best for: Enterprise-scale with custom requirements
V3.5 models include adaptive anchor optimization, component-aware generation, brand system enforcement, and custom fine-tuning. Contact your account manager for access and pricing.
Model IDs
| Model ID | Display Name | Plan Required |
|---|---|---|
| liff-v3-pro | LIFF V3 Pro | Pro |
| liff-v3-flash | LIFF V3 Flash | Starter |
| liff-v3.5-pro | LIFF V3.5 Pro | Enterprise |