FineEnvs/Qwen3.5-2B-opencode-standalone-RL
FineEnvs/Qwen3.5-2B-opencode-standalone-RL is a 2.3 billion parameter language model, fine-tuned from Qwen/Qwen3.5-2B by FineEnvs. It is specifically optimized for agentic data-analysis tasks using asynchronous GRPO with OpenCode, achieving a 29.8% pass@1 across four evaluation harnesses. This model excels in scenarios requiring automated data analysis and tool interaction, making it suitable for developing data agents.
Loading preview...
Qwen3.5-2B-opencode-standalone-RL Overview
This model is a 2.3 billion parameter language model, fine-tuned by FineEnvs from the Qwen/Qwen3.5-2B base model. It is specifically designed for agentic data-analysis tasks, utilizing asynchronous GRPO (TRL Async GRPO) with OpenCode for training. The model achieved a 29.8% pass@1 across four evaluation harnesses (OpenCode, Claude Code, Codex, Mini-SWE-Agent) on 250 fixed SmolDataEnvs test tasks.
Key Capabilities
- Agentic Data Analysis: Optimized for tasks requiring automated data interpretation and manipulation.
- Tool Interaction: Fine-tuned for effective interaction with external tools and environments, particularly OpenCode.
- Reinforcement Learning: Benefits from asynchronous GRPO training, focusing on binary correctness rewards.
- Performance: Achieves notable pass@1 scores on complex data analysis benchmarks, demonstrating proficiency in agent-based problem-solving.
Good For
- Developing data agents that can autonomously analyze and process data.
- Applications requiring automated code generation and execution within a sandbox environment.
- Research and development in reinforcement learning for language models in agentic settings.
- Scenarios where tool-use and interaction are critical for task completion, rather than just text generation.