AllModels平台目前已有五家独立的推理提供商支持“去审查”（abliterated）版本的DeepSeek模型。1 该平台通过将这些分散的资源整合至单一公共模型ID下，简化了开发者对新兴AI模型的集成与维护工作。1 具体而言，AllModels将所有abliterated模型整合到一个公共模型ID下，例如 'deepseek/deepseek-v4.1-flash-abliterated'。1
该统一路由服务完全兼容OpenAI的Chat Completions和Responses API格式。1 系统具备自动化故障转移机制：当某提供商容量不足或不可用时，流量自动切换到下一个可用路由。1 此外，网络搜索功能由AllModels托管，确保即使底层提供商不支持搜索，模型也能获得相同的搜索调用和URL引用。1
在请求处理策略上，AllModels实施了多模态请求智能分发逻辑：图像和多模态请求仅发送给支持这些功能的提供商，文本请求可使用所有符合条件的路由。1 该服务还提供最低成本路由功能，旨在简化开发者对新兴AI模型的集成与维护。1
AllModels has launched a unified model routing service that currently aggregates five independent inference providers offering support for the "abliterated" version of DeepSeek models 1. By consolidating these resources under a single public model ID, such as deepseek/deepseek-v4.1-flash-abliterated, the platform aims to simplify the integration and maintenance of emerging AI models for developers 1. The service is compatible with OpenAI’s Chat Completions and Responses API formats 1.
To ensure reliability and efficiency, the platform features an automated failover mechanism that redirects traffic to the next available route if a specific provider experiences capacity issues or becomes unavailable 1. It also employs intelligent distribution for multimodal requests, sending image-based queries only to providers that support these capabilities while routing text-only requests through all eligible pathways 1. Additionally, AllModels hosts a network search function, guaranteeing consistent search callouts and URL citations even when the underlying inference provider does not natively support web browsing 1.
评论
还没有评论，欢迎留下第一条。