A status quo investigation of large-language models for cost-effective computational fluid dynamics automation with OpenFOAMGPT

• OpenFOAMGPT extended to DeepSeek V3 and Qwen 2.5-Max for CFD automation • Reduces token cost by ∼100x compared to OpenAI o1 while preserving quality • Zero-shot prompts set up and debug diverse CFD cases without RAG support • Local 32B model on one GPU fails OpenFOAM syntax—fine-tuning still needed • Study charts path toward robust, low-cost LLM agents for industrial CFD We evaluated the performance of OpenFOAMGPT (GPT for generative pretrained transformers), which includes rating multiple large-language models. Some of the present models efficiently manage different computational fluid dynamics (CFD) tasks, such as adjusting boundary conditions, turbulence models, and solver configurations, although their token cost and stability vary. Locally deployed smaller models such as the QwQ-32B (Q4 KM quantized model) struggled with generating valid solver files for complex processes. Zero-shot prompts commonly fail in simulations with intricate settings, even for large models. Challenges with boundary conditions and solver keywords stress the need for expert supervision, indicating that further development is needed to fully automate specialized CFD simulations.