furiosa-ai/Qwen3-VL-32B-Instruct

VISIONConcurrent Unit Cost:2Model Size:33.4BQuant:FP8Context Size:32kTool Calling:SupportedPublished:Jun 9, 2026License:apache-2.0Architecture:Transformer Open Weights Featherless Exclusive Cold

Qwen3-VL-32B-Instruct is a 32-billion-parameter dense vision-language model from the Qwen3-VL series, developed by Qwen. It integrates a vision encoder with a dense transformer decoder, utilizing Interleaved-MRoPE and DeepStack for multimodal processing. This model excels at visual understanding tasks including OCR, document/chart analysis, spatial reasoning, and video comprehension, and natively supports tool calling. It is specifically optimized for deployment on FuriosaAI RNGD hardware via Furiosa-LLM.

Loading preview...

Qwen3-VL-32B-Instruct: A Dense Vision-Language Model

Qwen3-VL-32B-Instruct is a 32-billion-parameter dense vision-language model from the Qwen3-VL series, developed by Qwen. This model is designed to process both image and text inputs, generating text outputs. It incorporates Interleaved-MRoPE positional embeddings and DeepStack multi-level feature fusion to effectively handle images and videos alongside textual data.

Key Capabilities

  • Multimodal Understanding: Processes OpenAI-style multimodal chat messages, accepting image_url content parts alongside text.
  • Visual Task Proficiency: Capable of performing various visual understanding tasks, including OCR, document and chart analysis, spatial reasoning, and video comprehension.
  • Tool Calling: Natively supports tool (function) calling through the hermes tool-call parser, enabling interaction with external functions.
  • Hardware Optimization: Provided with a Furiosa Executable Bundle (FXB) for efficient execution on FuriosaAI RNGD hardware, utilizing a tensor-parallel size of 32 PEs across four RNGD cards.

Good For

  • Applications requiring advanced visual understanding and reasoning from images and videos.
  • Integrating AI models with external tools and functions for complex workflows.
  • Deploying high-performance vision-language models on FuriosaAI hardware for optimized inference.