zveloxy/Veloxy-Titan-32B-Vision

VISIONPricing:Input $1.6 / Cached $0.15 / Output $12Concurrent Unit Cost:2Model Size:27BQuant:FP8Context Size:32kTool Calling:SupportedPublished:Oct 8, 2026License:apache-2.0Architecture:Transformer0.0K Open Weights Featherless Exclusive Cold

Veloxy-Titan-32B-Vision by zveloxy is a 27 billion parameter multimodal model combining a 27-layer Vision Transformer with a 32B reasoning backbone, operating natively in bfloat16. It features dynamic multi-patch ViT encoding for high-DPI image resolution and a 131,072-token context window. This model excels at pixel-to-code generation from UI/UX mockups, analysis of technical schematics (e.g., PCB, cloud architecture), and complex document AI, making it suitable for applications requiring deep visual intelligence and structured output.

Loading preview...

Overview

Veloxy-Titan-32B-Vision is zveloxy's flagship multimodal model, integrating a 27-layer Vision Transformer (ViT) with a 32B reasoning backbone. This architecture allows it to process visual inputs, including UI/UX mockups, architectural schematics, and high-DPI scans, and generate outputs like production-ready code, structured JSON, or detailed technical analysis. The model operates natively in bfloat16 and supports a substantial context window of 131,072 tokens, enabling interleaved high-resolution visual tokens and long text.

Key Capabilities

  • Pixel-to-Code & UI/UX Reverse Engineering: Converts Figma exports, mobile screenshots, or web wireframes into clean, accessible code (e.g., React 19 / Next.js 15 with Tailwind CSS, Flutter, SwiftUI, HTML5 / CSS).
  • Technical Schematics & Circuit Diagram Analysis: Translates cloud architecture diagrams into Terraform / OpenTofu HCL code and analyzes electronics/PCB schematics, recognizing components and performing sanity checks. It also extracts database ERDs into Prisma, Drizzle, or raw SQL.
  • Document AI & Complex Table Intelligence: Accurately extracts data from multi-column financial statements and scanned legal forms into structured JSON or Markdown, preserving layout.
  • Visual Threat Modeling & Network Security: Audits network topology maps and firewall rule diagrams to detect vulnerabilities.

Good for

  • Developers and designers needing to convert visual mockups directly into functional code.
  • Engineers requiring automated analysis and code generation from technical diagrams like cloud architectures or PCB layouts.
  • Financial and legal professionals who need to extract structured data from complex, multi-column documents and scanned forms.
  • Security analysts performing visual threat modeling and network security audits.

Finetunes
1 models