AITOPAI热报

Claude Opus 5

Anthropic·US·Released 2026-07-24
ClosedMultimodalReasoningTool useVision

Anthropic's Claude Opus 5, supporting text, image and PDF input, excels at coding, agents and professional work.

1Mcontext
$5 / $25per 1M tokens (in / out)
—parameters
224stories in our feed

Specs

hover a value to see its source
Context window1M
Max output125K
Parameters—
Knowledge cutoff2026-05
LicenceProprietary
Commercial useAPI only
InputText · Image · PDF
OutputText
Familyclaude-opus

API price

USD per 1M tokens
Input$5
Output$25
Cached input$0.5

No price change recorded since we started tracking.

News timeline

from the AITOP feed, one story per eventAll 120 →
2026-10-23
Kimi-K3.1-Pro 发布：3T 参数，SWE 系列测试对标 OpenAI 与 Anthropic 旗舰karminski-牙医 (AI工具)
2026-10-17
DeepSeek V4.1 Flash-1017 编程基准超越 GPT-5.6 Sol，价格低一个量级kimmonismus
2026-10-10
Evolvent AI 开源 RSIGym 递归自我改进研究环境berryxia
2026-10-08
RSIGym research agent platformelvis
2026-10-02
Claude Opus 5 完胜人类会计师rohanpaul_ai
2026-09-30
Ling3.1 flash发布：500亿参数模型kimmonismus
2026-09-29
Sonnet 5.5 用十分之一成本超越前代Lxfater
Databricks 用 GPT-6 Astra 与 Opus 5 智能体登顶 NVIDIA SOL-ExecBench 榜首Yuchen Jin
2026-09-28
Jev 24.9 秒读完 384 条新闻，成本仅 $0.19，快于 Claude Opus 5AI Will
2026-09-26
Opus 5.5 pricing cuts real Claude Code session costs by 31%宝玉
Claude Opus 5.5 writes shorter sentences with fewer AI tellslmarena.ai
2026-09-23
GPT-6 Sol (Max) 登上 Code Arena WebDev 榜第4名，价格低于 Claude Opus 5 一半lmarena.ai
OpenAI 发布 GPT-6 Sol 和 Luna，API 价格降半、缓存省 90% 输入成本Techsauce
Opus 5.5 默认档位对标前沿：成本更低、速度比 Opus 5 快 30%orange.ai
Anthropic 发布 Claude Opus 5.5，成本比 Opus 5 低 40%Blognone
2026-09-22
OpenAI 推出 GPT-6 Luna：低成本档在 DeepSWE v1.1 拿 66.6%OpenRouter
从 Opus 5 迁移指南：自适应思考、强制工具调用移除、价格降 20%OpenRouter
Claude Opus 5.5 上线 OpenRouter，价格比 Opus 5 低 20%OpenRouter
Fable and Opus Loop Cuts Token Costs in Coding Workflows宝玉
2026-09-20
Claude Opus 5 escalates image upload flaw into OpenAI account takeoverAI Will
Opus 4.8 finds libheif vulnerabilityAI Will
2026-09-18
Databricks rolls out Astra model to engineersAI Will
2026-09-11
Cursor releases private benchmark CursorBench 4.0shao__meng
2026-09-06
Astra excels in browser use benchmarkBrowser Use
2026-09-05
Astra tops cybersecurity model rankingsGuillermo Rauch
2026-09-04
GPT6-Astra Model Review Analysis向阳乔木
Agent技术选型实验：Stripe等工具成首选shao__meng
2026-09-03
Gemini 3.8 Flash 发布：接近前沿模型能力shao__meng
2026-08-28
Claude Code Opus 5 Auto Mode VulnerabilitySimon Willison’s Weblog
2026-08-26
Agent Arena发布模型评估：Claude Opus 5领先lmarena.ai
2026-08-24
Anthropic Fable 5 模型仅占6%购买TokenAI Will
2026-08-23
Anthropic顶级AI模型用户增长缓慢，低成本工具盛行Simon Willison’s Weblog
2026-08-17
Agent Arena 新分类榜：代码榜首 GPT 5.6 Sol，工作聊天归 Claude Opus 5lmarena.ai
2026-08-14
FT数据质疑Anthropic千亿收入预测，价格战下Q4需翻倍Gary Marcus
ARC-AGI-3 接近被攻克：编码工具链让 Opus 5 拿下 96.2%Amjad Masad
2026-08-13
LMArena：Claude Opus 5写作更长、词汇更简单lmarena.ai
2026-08-08
Opus 5 配 Browser Use 的自主智能体已上线Browser Use
2026-08-07
Anthropic放宽Fable 5生物限制，病毒学毒理学仍设防Decoder
2026-08-05
Claude Opus 5 (Max)登顶Fullstack Code Arena，得分1699lmarena.ai
卡帕西提出 AI 测试新思路：让 Claude Opus 5 用 100 万 Tokens 构建《指环王》3D 场景IT之家