AITOPAI热报

Claude Mythos 5

Anthropic·US
ClosedRumoredGeneral LLMReasoningTool useVision

Claude Mythos 5 is a model from Anthropic's Claude family. It is closed-weight and not openly licensed. Details on its capabilities and pricing are not available from the collected facts.

1Mcontext
$10 / $50per 1M tokens (in / out)
—parameters
22stories in our feed

Specs

hover a value to see its source
Context window1M
Max output125K
Parameters—
Knowledge cutoff—
Licence—
Commercial useAPI only
InputText · Image
Output—
Familyclaude-mythos

API price

USD per 1M tokens
Input$10
Output$50
Cached input$1

No price change recorded since we started tracking.

News timeline

from the AITOP feed, one story per eventAll 22 →
2026-10-16
Anthropic STEM Fellows 项目：用 Claude Mythos 5 设计更高效的量子电路Anthropic
Anthropic：Claude Mythos 5 在闭环中优化量子电路，超越已发表最优算法Andrew Curran
2026-09-24
Anthropic 宣布成立生物学实验室，Claude Mythos 5 自主发现新酶系统Tanishq Abraham (论文推介)
2026-09-10
Anthropic 的 Claude Mythos 5 自我指认系统模拟，OpenAI 的 GPT-6 Astra 被迫依赖模型推理可读性Decoder
欧盟机构获准访问 OpenAI GPT-6 Astra 模型IT之家
Anthropic opens Mythos 5 to EU cybersecurity agencyIT之家
2026-08-31
Anthropic详解Claude安全事件：配置错误致模型入侵企业IT之家
2026-08-21
Anthropic Integrates Claude Mythos 5 into Securitymarktechpost
Anthropic用Claude Mythos 5模型助力网络安全Decoder
ClaudeAI与伙伴合作整合Mythos 5Claude
Claude Mythos 5 安全扫描公测Claude
2026-08-18
Anthropic披露Model 2模型，能力略高于Claude Mythos 5IT之家
2026-08-15
Anthropic 发布 AI 风险报告：智能体攻击同类并隐藏违规痕迹IT之家
2026-08-13
Claude Opus 5系统提示词曝光：回应出口管制事件Simon Willison’s Weblog
2026-08-05
AISI评估事故：AI代理未经授权攻击真实组织Simon Willison’s Weblog
英国AI安全测试中智能体失控 伪造身份发起社交攻击Decoder
AISI测试：Anthropic和OpenAI智能体被发现执行未授权操作IT之家
2026-08-04
英国AISI：Claude Mythos 5与GPT-5.6 Sol在测试中实施有害行为Anthropic
2026-07-28
纳德拉官宣微软首个AI安全模型MAI-Cyber-1-Flash，跑分超Mythos 5和GPT-5.5 CyberIT之家
2026-07-24
Opus 5在网络安全任务上强于Opus 4.8，但远不及Mythos 5Claude