ricdomolm/mini-coder-1.7b
ricdomolm/mini-coder-1.7b is a 1.7 billion parameter model distilled from Qwen 3 Coder 30B A3B, specifically designed for software engineering tasks. It demonstrates strong performance on SWE-bench Verified Bash only, outperforming larger models like SWE-agent-LM 7B. This model is optimized for code generation and problem-solving within a software development context, making it suitable for integration into agentic frameworks.
Loading preview...
Overview
ricdomolm/mini-coder-1.7b is a compact yet powerful 1.7 billion parameter model, distilled from the larger Qwen 3 Coder 30B A3B. It is specifically engineered for software engineering tasks, showcasing performance that "punches well above its weight" on benchmarks like SWE-bench Verified Bash only.
Key Capabilities
- Code Generation & Problem Solving: Excels in software engineering tasks, as evidenced by its strong performance on SWE-bench.
- Efficient Performance: Outperforms larger models such as SWE-agent-LM 7B on specific code-related benchmarks.
- Agentic Integration: Designed to work seamlessly with lightweight agentic frameworks like
mini-swe-agent, facilitating RL fine-tuning. - Resource-Friendly Fine-tuning: Can be post-trained on a single 80GB GPU or smaller, making it accessible for developers with limited hardware.
- Dense Model Architecture: Benefits from a more mature fine-tuning ecosystem compared to Mixture-of-Experts (MoE) models.
Good For
- Automated Software Engineering: Ideal for tasks requiring automated code fixes, bug resolution, and general software development assistance.
- Resource-Constrained Environments: Suitable for developers and researchers looking for a capable code model that can be fine-tuned on consumer-grade GPUs.
- Agentic Workflows: Best utilized within agentic scaffolding like
mini-swe-agentfor generating SWE-bench trajectories and similar automated development processes.
Top 3 parameter combinations used by Featherless users for this model. Click a tab to see each config.