本文作者来自马里兰大学的 CASE (Collaborative, Automated, Scalable, and Efficient Intelligence) Lab，主要参与者为博士生孙国恒与王子瑶，指导教师为李昂教授。 近年来，大型语言模型（LLM）在处理复杂任务方面取得了显著进展，尤其体现在多步推理、工具调用以及多 ...

摘要：在 AI 调用量激增的今天，为什么你的 LLM API 账单越来越贵？本文揭秘 AI 大模型 计费中不为人知的“汇率”与“隐形”损耗。通过深度解析 OpenRouter 与 n1n.ai 的定价策略，助你以此为鉴，实现 AI 大模型 成本降低 85% 的目标。 2025年，随着 GPT-4o 和 Claude 3.5 的 ...

谷歌云推出了 Model Armor 的 公开预览版，这是一个集成到 Apigee API 管理平台的原生大语言模型（LLM）治理框架。一篇社区帖子中详细介 绍了 Model Armor，它引入了针对 LLM 特定策略的即开即用执行，例如提示验证、输出过滤和 API 层的 token 级控制。 Model Armor 直接在 ...

Protein large language model (LLM) designed to help enterprises accelerate drug development coming to Google Cloud's Vertex AI Model Garden soon; one of the first-of-its-kind in the industry Model API ...

LiteLLM allows developers to integrate a diverse range of LLM models as if they were calling OpenAI’s API, with support for fallbacks, budgets, rate limits, and real-time monitoring of API calls. The ...

Claude quickly became the developer’s top choice for code generation, capturing 42% market share, more than double OpenAI’s (21%). In just one year, Claude helped transform a single-product space ...

API security provider Cequence has added new large language model (LLM) threat detection and management capabilities along with some fresh integrations for API discovery on its Unified API protection ...

Earlier this week Anthropic surprise the AI community by releasing three new AI models making up the Claude 3 family. The three different-sized models: Haiku, Sonnet, and Opus are vision language ...

Sophisticated "LLMjacking" operations have obtained stolen access to DeepSeek models, just weeks after their public release. Most recently, researchers from Sysdig observed hyperactive LLMjacking ...