covaga/eplan-2027-9b-merged
covaga/eplan-2027-9b-merged is a 9 billion parameter Qwen3.5-based model, fine-tuned by covaga for EPLAN Electric P8 2027 API assistance. This BF16 merged model, with a 32768 token context length, excels at generating C# scripts for the EPLAN scripting engine, diagnosing errors, and answering API-specific questions. It is bilingual, supporting both English and Spanish, and is designed to refuse nonexistent APIs.
Loading preview...
EPLAN Electric P8 2027 API Assistant
covaga/eplan-2027-9b-merged is a specialized 9 billion parameter model based on Qwen3.5, specifically fine-tuned for EPLAN Electric P8 2027 API interactions. This BF16 merged model integrates the eplan-2027-lora-v3 adapter, allowing direct loading without PEFT.
Key Capabilities
- EPLAN API Expertise: Answers questions about the EPLAN Electric P8 2027 API (
Eplan.EplApi.*). - C# Script Generation: Creates C# scripts for the EPLAN scripting engine.
- Error Diagnosis: Explains compile errors (e.g., CS0234) and runtime issues.
- API Refusal: Designed to refuse (most of the time) queries for nonexistent APIs, reducing fabrication.
- Bilingual Support: Operates effectively in both English and Spanish.
Performance & Training
On a curated 44-question evaluation, this model achieved 63% valid identifiers and significantly reduced invented IDs to 47, compared to the base Qwen3.5-9B's 43% valid identifiers and 230 invented IDs. It was trained on the covaga/eplan-2027-api-qa dataset, comprising ~19k Q&A pairs generated from official EPLAN 2027 API documentation, plus 246 curated examples for refusal and error diagnosis.
Use Cases
This model is ideal for developers and engineers working with EPLAN Electric P8 2027, providing targeted assistance for API usage, script development, and troubleshooting. For enhanced accuracy, especially regarding API refusal, pairing with retrieval-augmented generation (RAG) systems is recommended.