OrionLLM/Terminus-Qwen3-8b
OrionLLM/Terminus-Qwen3-8b is a model developed by OrionLLM, fine-tuned for terminal agentic tasks. It is specifically optimized for performance in environments like Terminal-Bench 2.0 and SWE-Bench, and efficient for use with Codex and OpenCode. This model excels at terminal-based reasoning workflows and software engineering tasks, demonstrating significantly improved benchmark scores over its base model.
Loading preview...
Overview
OrionLLM/Terminus-Qwen3-8b is a specialized model developed by OrionLLM, focusing on enhancing performance for terminal agentic tasks. It is built upon the Qwen3-8B architecture and has been specifically trained using the OrionLLM/OpenAgentInstruct dataset.
Key Capabilities
- Terminal Agentic Tasks: Designed to excel in complex terminal-based reasoning workflows.
- Software Engineering: Optimized for software engineering tasks, including those found in SWE-Bench.
- Tool Usage: Efficient for integration and localization within environments like Codex and OpenCode.
Performance Highlights
Terminus-Qwen3-8b demonstrates notable improvements over its base model, Qwen3-8B, and even larger models like Qwen3-32B, on specific benchmarks:
- Achieves 4.9 on Terminal-Bench 2.0, significantly higher than Qwen3-8B's 0.0 and Qwen3-32B's 1.9.
- Scores 15.7 on SWE-Bench Verified, outperforming Qwen3-8B's 0.7 and Qwen3-32B's 5.7.
When to Use This Model
This model is ideal for developers and researchers working on:
- Automating tasks within terminal environments.
- Developing and testing agentic systems that interact with command-line interfaces.
- Software development workflows requiring robust code generation and problem-solving capabilities within a terminal context.