vectionlabs/Salience-1-9B

Hugging Face
VISIONConcurrent Unit Cost:1Model Size:8BQuant:FP8Context Size:32kTool Calling:SupportedPublished:Jun 10, 2026License:apache-2.0Architecture:Transformer0.0K Open Weights Featherless Exclusive Warm

Salience 1 (9B) by Vection Labs is a dense, 9-billion-parameter multimodal vision-language model built on the Qwen3-VL architecture, featuring a Qwen3-8B language model and a native vision encoder. It is engineered for hard, practical work, excelling in code generation, agentic tasks, multi-step mathematical reasoning, and visual understanding over images and video. The model supports an extended context window of up to 1 million tokens via interleaved multimodal RoPE, making it suitable for processing entire code repositories or long documents.

Loading preview...

Salience 1 (9B) Overview

Salience 1 (9B) is a 9-billion-parameter multimodal reasoning model developed by Vection Labs, designed for demanding technical applications. It builds upon the Qwen3-VL architecture, integrating a Qwen3-8B language model with a native vision encoder. This model is specifically optimized for code and agentic/tool use, while maintaining strong capabilities in deep reasoning and multimodal perception.

Key Capabilities

  • Code & Agentic First: Tuned to produce runnable code, facilitate repo-scale edits, and generate well-formed tool calls for agentic workflows.
  • Deep Reasoning: Provides structured, inspectable chains of thought for complex mathematical problems, logic, and code analysis.
  • Genuinely Multimodal: Processes both images and video as first-class inputs, enabling visual understanding over diagrams, UI screenshots, and short video clips.
  • Long Context: Features an extensive context window of up to 1 million tokens through interleaved multimodal RoPE, allowing for comprehensive analysis of large documents or codebases.
  • Efficiency: Designed for fast inference on modest hardware, running on 2x T4 GPUs without GGUF (fp16 sharded) or 4-bit on a single T4, and supports speculative decoding for speedups.

Intended Use Cases

Salience 1 is particularly well-suited for:

  • Code generation, explanation, debugging, and review.
  • Agentic and tool-using workflows requiring structured outputs.
  • Step-by-step mathematical and quantitative reasoning.
  • Visual question answering and understanding of documents, diagrams, and charts.
  • Video understanding over short clips and long-document analysis.