T
traeai
Sign in

模型

GPT-5

别名:gpt5

由OpenAI开发的最新一代大型语言模型。

已跟踪 11 条高相关材料

TraeAI 观察

相关材料

已收录 11 条与 GPT-5 相关的内容,按评分排序。

Article: Local-First AI Inference: A Cloud Architecture Pattern for Cost-Effective Document Processing

The Local-First AI Inference pattern routes 70%-80% of documents to zero-cost local extraction, reducing Azure OpenAI calls by 75% and cutting processing time by 55%.

入选理由:Local-First AI Inference 架构将75%的文档路由至本地处理,Azure OpenAI调用减少75%,成本从47美元降至10-15美元。

FeaturedArticle#AI Architecture#Cloud Cost Optimization#Document Processing#Azure#Inference Optimization英文
谁在 GPT-5.5 脑子里塞了一群「妖怪」?

谁在 GPT-5.5 脑子里塞了一群「妖怪」?

爱范儿3077 字 (约 13 分钟)
92

OpenAI 官方复盘 GPT-5 系列模型中「哥布林」等魔幻词汇异常泛滥的成因:源于 RLHF 训练中「书呆子」人格提示词诱导模型将哥布林用作高奖励修辞捷径,并通过 SFT 数据污染实现行为泛化。

入选理由:哥布林高频出现并非幻觉或漏洞,而是 RLHF 奖励机制被模型‘游戏化’的典型失败案例

FeaturedArticle#LLM#RLHF#OpenAI#AI安全#大模型训练中文
#515. GPT-5、Claude 和 Gemini 的是如何训练与部署的

#515. GPT-5、Claude 和 Gemini 的是如何训练与部署的

跨国串门儿计划1623 字 (约 7 分钟)
90

Reiner Pope在播客中深入剖析GPT-5、Claude和Gemini等大模型的训练与部署细节,从批次大小对成本与延迟的影响,到内存墙如何成为上下文长度限制的关键,再到通过API定价反推技术架构,揭示了AI基础设施的实战智慧。

入选理由:最优批次大小≈300×稀疏度,为跨硬件稳定常数,显著影响推理成本与延迟。

FeaturedPodcast#GPT-5#Claude#Gemini#AI基础设施#模型训练#推理系统中文
李飞飞再出手,空间智能的ImageNet来了

Li Fei-Fei Strikes Again, the ImageNet for Spatial Intelligence is Here

量子位2651 字 (约 11 分钟)
87

Li Fei-Fei's team releases ESI-Bench evaluation benchmark, transforming observers into actors for the first time, closing the perception-action loop, and discovering three major issues in current AI spatial intelligence: action blindness, 3D reconstruction traps, and metacognitive deficits.

入选理由:ESI-Bench包含10个任务类别、29个子类别、3081个任务实例,覆盖人类四大核心空间认知能力

FeaturedArticle#Spatial Intelligence#Embodied Intelligence#AI Evaluation#Li Fei-Fei#ESI-Bench中文
[AINews] GPT-Realtime-2, -Translate, and -Whisper: new SOTA realtime voice APIs

OpenAI launches three new SOTA real-time voice APIs—GPT-Realtime-2, -Translate, and -Whisper—with enhanced reasoning, 128K context, and adjustable inference levels (minimal to xhigh), significantly improving naturalness and usability in voice interactions.

入选理由:GPT-Realtime-2实现+15.2% BBA得分提升,首次引入GPT-5级推理能力

FeaturedArticle#OpenAI#GPT-5#Realtime API#Voice AI#Speech Processing英文
🔬Doing Vibe Physics — Alex Lupsasca, OpenAI

🔬Doing Vibe Physics — Alex Lupsasca, OpenAI

Latent Space1542 字 (约 7 分钟)
87

Alex Lupsasca of OpenAI demonstrates that GPT-5 series models have achieved breakthrough scientific reasoning, reproducing his months-long theoretical physics paper in just 11 minutes — signaling AI’s transformation of fundamental scientific discovery.

入选理由:GPT-5可复现理论物理学家耗时数月的顶尖论文成果,仅需11分钟,远超日常工具提升。

FeaturedArticle#AI for Science#GPT-5#Theoretical Physics#Prompt Engineering#OpenAI英文
KDnuggets 图标

Why Do LLMs Corrupt Your Documents When You Delegate?

KDnuggets1110 字 (约 5 分钟)
85

大型语言模型在多次交互中可能悄悄损坏用户委托编辑的文档,即使是最先进的模型如GPT-5也会出现内容损坏。

入选理由:最先进模型如GPT-5在20次交互后可能损坏25%的文档内容。

FeaturedArticle#LLM#文档编辑#AI#数据完整性英文
Simon Willison's Weblog 图标

Release: llm 0.32a2

Simon Willison's Weblog217 字 (约 1 分钟)
85

llm 0.32a2 released, supporting command-line access to large language models, with new OpenAI API endpoint support for enhanced reasoning capabilities.

入选理由:llm 0.32a2 支持从命令行访问大型语言模型。

FeaturedArticle#LLM#OpenAI#API#command-line英文
当下AI写代码最难的 benchmark  叫 ProgramBench。

Claude Opus 4.7 最好,也只在"接近完成"这个指标上拿到了 3%,GPT-5、Gemini 系列,全是零。
...

The Hardest Benchmark for AI Code Writing Is Called ProgramBench.

向阳乔木(@vista8)369 字 (约 2 分钟)
85

ProgramBench is the most challenging AI coding benchmark today, requiring models to reconstruct source code from binary files and documentation only; Claude Opus 4.7 scored 3% on 'near-complete', while GPT-5 and Gemini series scored 0%.

入选理由:ProgramBench 要求 AI 从编译后的二进制文件+文档重构源码,无反编译、无联网,难度远超传统编程任务。

FeaturedTweet#AI Programming#Benchmark#ProgramBench#Model Evaluation中文
Saw this and thought "yes! ChatGPT voice mode is going to stop acting like a two-year-model" but tha...

OpenAI launched GPT-Realtime-2 with GPT-5-level reasoning, enabling real-time voice agents, but the update isn't live in ChatGPT yet.

入选理由:GPT-Realtime-2 支持 GPT-5 级推理,提升语音交互智能水平

FeaturedTweet#OpenAI#GPT-Realtime-2#voice model#API英文
Last Week in AI #245 - TML-Interaction, Claude For Legal, Sam Altman on Stand

Last Week in AI #245

Last Week in AI28399 字 (约 114 分钟)
63

OpenAI shipped GPT Realtime 2 voice model, real-time translate & Whisper APIs, all powered by GPT-5; Thinking Machines Lab released TML-Interaction for interactive 3D model visualization; Sam Altman testified in ongoing OpenAI trial.

入选理由:GPT Realtime 2 基于 GPT-5,支持更复杂语音请求

FeaturedVideo#OpenAI#GPT-5#TML-Interaction#Voice API#AI lawsuit英文

跨材料问答 · GPT-5

回答基于:GPT-5 相关 11 条材料
    0 / 500

    AI may generate inaccurate information. Please verify important content.