osunlp/QUEST-2B
osunlp/QUEST-2B is a 2.3 billion parameter SFT (Supervised Fine-Tuned) model from the QUEST family, based on the Qwen3.5 architecture. Developed by osunlp, this dense model is designed as a general-purpose deep research agent, excelling in objective tasks and reasoning-heavy benchmarks. It features a 32768 token context length and is optimized for tasks like web browsing, information retrieval, and complex research queries.
Loading preview...
Overview
osunlp/QUEST-2B is a 2.3 billion parameter SFT (Supervised Fine-Tuned) model, part of the broader QUEST family developed by osunlp. Built on the Qwen3.5 architecture, this model is specifically designed to function as a general-purpose deep research agent, capable of handling complex information retrieval and reasoning tasks. It is one of several models in the QUEST family, which ranges from 2B to 35B parameters, offering various checkpoints including SFT, MT, MT+SFT, and RL versions.
Key Capabilities
- Deep Research Agent: Optimized for tasks requiring extensive information processing and synthesis.
- Objective Task Performance: Demonstrates strong performance on reasoning-heavy objective benchmarks, with scores such as 28.0 on BrowseComp, 8.8 on Mind2Web 2, 30.3 on HLE, and 72.8 on GAIA.
- Extended Context: Features a 32768 token context length, suitable for processing longer documents and complex queries.
Good For
- Automated Web Browsing and Search: Achieves 52.6 on BrowseComp-Plus and 40.9 on WideSearch, indicating proficiency in navigating and extracting information from the web.
- Complex Reasoning and Research: Its strong benchmark results on tasks like GAIA and DeepResearch Bench (21.0) suggest suitability for applications requiring advanced analytical capabilities.
- Integration into Research Workflows: Can be used as a component in systems requiring automated data collection, analysis, and summarization.