一篇读懂 Token 计费：上下文、输入输出与一份成本优化清单
「这个应用每月的模型账单会是多少」——回答这个问题需要真正理解 token 计费的口径：输入输出分价、上下文累积、缓存折扣、以及隐藏的成本大头。本篇拆解计费的完整口径、一笔真实对话的成本拆解、以及按效排序的成本优化清单——把「token 焦虑」变成「token 预算学」。
阅读全文 →
全部文章
ARCHIVE 共 425 条 刚刚，GPT-6 新模型掀桌！「白菜价」杀进 DeepSeek 腹地
DeepSeek 快醒醒，ChatGPT 要打到家门口了。
就在刚刚，OpenAI 正式发布 GPT-6 Sol 和 GPT-6 Luna，API 价格降至上一代的一半左右。其中，Luna 每百万 Token 输入 0.1 美元、输出 0.…
Training Object Permanence in World Models
Object permanence and solidity are hallmarks of human cognitive priors. Recent studies show that video generation models…
Anthropic 发布 Claude Opus 5.5 模型：性能媲美 Fable 5.1，运行成本比 Opus 5 低 40%
IT之家 9 月 23 日消息，Anthropic 今日正式发布了 Claude Opus 5.5 模型，这是其全新 Claude 5.5 家族的首个模型，**宣称在大多数工作上表现可媲美 Claude Fable 5.1，运行成本比 Op…
Lawsuit demands OpenAI pay for new school after ChatGPT used in shooting
The families of victims killed in one of the deadliest mass shootings in Canada’s history have already sued OpenAI over …
消息称 DeepSeek 本周参加联合国安理会，讲解 AI 安全风险
IT之家 9 月 22 日消息，据路透社今日援引知情人士消息，中国人工智能初创公司 DeepSeek（深度求索）将在本周参加联合国安理会，**讲解 AI 安全风险**。与此同时，世界各国领导人将齐聚纽约，参加一年一度的联合国大会。
IT之家…
PACT: From Credit Assignment to Critic Alignment
Reinforcement learning has become a central component of large language model (LLM) post-training, yet token-level credi…
Disaggregated Quantization: Specializing LLM Prefill and Decode
Prefill and decode reward different approaches to quantization: low-precision arithmetic accelerates prompt processing, …
联合国秘书长古特雷斯呼吁全球监管人工智能，避免出现杀人机器人
IT之家 9 月 22 日消息，据彭博社报道，联合国秘书长进一步呼吁对人工智能实施监管，以避免出现企业权力泛滥、杀人机器人横行的反乌托邦未来。
安东尼奥 · 古特雷斯（Antonio Guterres）在联合国大会上向各国领导人发表卸任前最…
Don’t be fooled by this summer of AI hype
It’s been a busy few months for AI hype. At the end of April, Anthropic claimed that its model Claude Mythos is better a…
Alibaba Qwen Releases Qwen-Image-2.1: A 7B Open-Weight Model for Image Generation and Editing
Alibaba's Qwen team has released Qwen-Image-2.1, a 7B diffusion transformer that handles text-to-image generation, multi…
Xiaomi's affordable flagship AI leads the open models, and Anthropic says Claude helped get it there
Sep 22, 2026
Xiaomi
Key Points
Xiaomi's new MiMo-V2.6-Pro model leads current rankings of open AI models while costing f…
Flash-dLLM: IO-Aware KV Caching and Parallel Decoding for Fast, Memory-Efficient Diffusion LLMs
Diffusion Large Language Models (dLLMs) have recently emerged as a promising alternative to autoregressive LLMs by enabl…