OrionLLM/Terminus-Qwen3-8b
OrionLLM/Terminus-Qwen3-8b is a model based on the Qwen3 architecture, specifically trained by OrionLLM for terminal agentic tasks. It demonstrates enhanced performance on benchmarks like Terminal-Bench 2.0 and SWE-Bench Verified, making it efficient for environments such as Codex and OpenCode. This model is optimized for terminal-based reasoning workflows, software engineering, and advanced tool usage.
Loading preview...
Overview
OrionLLM/Terminus-Qwen3-8b is a specialized model developed by OrionLLM, building upon the Qwen3 architecture. Its primary focus is on terminal agentic tasks, aiming to improve performance in complex interactive environments. The model was trained using the proprietary OrionLLM/OpenAgentInstruct dataset, specifically curated to enhance its capabilities in these domains.
Key Capabilities
- Terminal Agentic Tasks: Designed for high efficiency in terminal-based reasoning workflows.
- Software Engineering: Optimized for tasks related to software development, including code generation and problem-solving.
- Tool Usage: Engineered to integrate seamlessly with environments like Codex and OpenCode, facilitating advanced tool interaction.
Performance Highlights
Terminus-Qwen3-8b shows significant improvements over base Qwen3 models on agentic benchmarks:
- Achieves 4.9 on Terminal-Bench 2.0, compared to 0.0 for Qwen3-8B and 1.9 for Qwen3-32B.
- Scores 15.7 on SWE-Bench Verified, substantially higher than 0.7 for Qwen3-8B and 5.7 for Qwen3-32B.
Good For
- Developers and researchers working on automated terminal interactions.
- Applications requiring robust performance in software engineering tasks and code environments.
- Use cases demanding enhanced agentic reasoning and tool integration.