讯飞星辰MaaS

讯飞星辰MaaS Key Features

• Zero-Code Model Fine-Tuning

  Fine-tune large models without coding by uploading datasets, selecting models, and configuring parameters

• 20+ Industry Model Library

  Aggregates 20+ quality models including Spark, Llama, and DeepSeek, covering multiple industry scenarios

• Multi-Model Batch Inference

  Supports up to for simultaneous online batch inference, facilitating fine-tuning comparison and model selection

• Training Monitoring & Evaluation

  Real-time Loss curve display for training monitoring, with judge scoring model to evaluate fine-tuning effectiveness

讯飞星辰MaaS Key Advantages

• Zero-code fine-tuning requires no programming; upload dataset + select model + configure params to complete fine-tuning
• Up to for simultaneous batch inference enables quick fine-tuning comparison
• OpenAI protocol compatible with function_call fine-tuning support for low migration cost
• New users get 10 hours free GPU resources, sufficient for initial fine-tuning experiments

讯飞星辰MaaS Use Cases

• Enterprise Custom Models

  Enterprises upload industry data to fine-tune custom models, improving model performance in specific business scenarios

• Model Performance Comparison

  Use batch inference to compare multiple fine-tuned models simultaneously and select the best option

Frequently Asked Questions