ClaudeClaude (Anthropic Messages)
apinebula.ai
Claude Sonnet 4.6 (claude-sonnet-4-6) · 标准模式 · 08/29 11:40 · 耗时 33s
各项指标正常,Claude Sonnet 4.6 通过验证
81
通过
首 token 1.51s · 往返 2.18s
通过协议字段合规响应字段与官方 Messages API 规范一致权重 10 · 100›
响应字段与官方 Messages API 规范一致
{
"status": 200,
"id": "msg_011CeWyYFGPUPupLfirH6H47",
"model": "claude-sonnet-4-6",
"stop_reason": "end_turn",
"usage": {
"input_tokens": 5,
"cache_creation_input_tokens": 0,
"cache_read_input_tokens": 0,
"cache_creation": {
"ephemeral_5m_input_tokens": 0,
"ephemeral_1h_input_tokens": 0
},
"output_tokens": 1,
"service_tier": "standard",
"inference_geo": "global"
},
"topKeys": [
"model",
"id",
"type",
"role",
"content",
"stop_reason",
"stop_sequence",
"stop_details",
"usage",
"context_management"
],
"latencyMs": 2177,
"text": "OK"
}耗时 2.2s · 严重度 info
通过消息 ID 规范消息 ID 符合官方格式且每次唯一权重 4 · 100›
消息 ID 符合官方格式且每次唯一
{
"ids": [
"msg_011CeWyYFGPUPupLfirH6H47",
"msg_011CeWyYNA8DihZkY3et6oPa"
]
}耗时 1.6s · 严重度 info
失败思维签名加密验证首轮签名形态为 Anthropic 官方,但篡改后的签名回传也被接受:中转站在转发时改写了对话历史(剥离/替换 thinking 块),签名验证链在中转站被切断——多账号号池轮换的典型做法。上游可能是官方账号,但你无法验证后续请求是否仍由官方处理权重 25 · 40›
首轮签名形态为 Anthropic 官方,但篡改后的签名回传也被接受:中转站在转发时改写了对话历史(剥离/替换 thinking 块),签名验证链在中转站被切断——多账号号池轮换的典型做法。上游可能是官方账号,但你无法验证后续请求是否仍由官方处理
{
"blockTypes": [
"thinking",
"tool_use"
],
"stop_reason": "tool_use",
"signatureLength": 436,
"signaturePrefix": "EsECCosBCBAY",
"thinkingPreview": "The user wants me to check the current weather in Tokyo using the get_weather to…",
"thinkingMode": "enabled",
"signatureLooksOfficial": true,
"roundTrip": {
"validStatus": 200,
"tamperedStatus": 200,
"tamperedError": "{\"model\":\"claude-sonnet-4-6\",\"id\":\"msg_011CeWyYeVPRP91GUCL4z2cM\",\"type\":\"message\",\"role\":\"assistant\",\"content\":[{\"type\":\"thinking\",\"thinking\":\"The user wants me…"
}
}耗时 4.7s · 严重度 major
通过流式响应合规SSE 事件序列完整,符合官方规范权重 8 · 100›
SSE 事件序列完整,符合官方规范
{
"status": 200,
"ttftMs": 1509,
"eventCount": 8,
"sequence": [
"message_start",
"content_block_start",
"ping",
"content_block_delta",
"content_block_delta",
"content_block_stop",
"message_delta",
"message_stop"
],
"startId": "msg_011CeWyYqLVzzV8o3jzLpFvv",
"startUsage": {
"input_tokens": 13,
"cache_creation_input_tokens": 0,
"cache_read_input_tokens": 0,
"cache_creation": {
"ephemeral_5m_input_tokens": 0,
"ephemeral_1h_input_tokens": 0
},
"output_tokens": 1,
"service_tier": "standard",
"inference_geo": "global"
},
"deltaUsage": {
"input_tokens": 13,
"cache_creation_input_tokens": 0,
"cache_read_input_tokens": 0,
"output_tokens": 9
}
}耗时 1.9s · 严重度 info
通过Tool Use 能力Tool Use 正常,参数与 stop_reason 均符合规范权重 8 · 100›
Tool Use 正常,参数与 stop_reason 均符合规范
{
"stop_reason": "tool_use",
"blockTypes": [
"tool_use"
],
"toolId": "toolu_01Q7fApEmroFRFGzkbBwZAvg",
"input": {
"city": "Tokyo"
}
}耗时 2.1s · 严重度 info
跳过Assistant 预填充该模型不支持 assistant 预填充(Claude 5 代已移除此特性),跳过权重 6 · —›
该模型不支持 assistant 预填充(Claude 5 代已移除此特性),跳过
耗时 1.2s · 严重度 info
通过stop_sequences 行为stop_sequence 命中并正确回填权重 4 · 100›
stop_sequence 命中并正确回填
{
"stop_reason": "stop_sequence",
"stop_sequence": "FIVE",
"text": "ONE TWO THREE FOUR "
}耗时 3.0s · 严重度 info
通过Token 用量与注入用量稳定 (5 tokens),无注入痕迹权重 8 · 100›
用量稳定 (5 tokens),无注入痕迹
{
"input_tokens": [
5,
5
],
"output_tokens": [
1,
1
],
"expectedInputApprox": "10-16"
}耗时 1.7s · 严重度 info
失败隐藏 System Prompt 探测模型复述出了隐藏 system prompt,中转站在注入指令权重 5 · 20›
模型复述出了隐藏 system prompt,中转站在注入指令
{
"reply": "I'm Claude Code, Anthropic's official CLI for Claude. I won't reproduce my setup text, but I'm glad to help with what you're working on."
}耗时 2.2s · 严重度 major
通过身份与确定性身份为 Anthropic,temperature=0 输出稳定权重 5 · 100›
身份为 Anthropic,temperature=0 输出稳定
{
"who": "Anthropic",
"math": [
"391",
"391"
],
"models": [
"claude-sonnet-4-6",
"claude-sonnet-4-6",
"claude-sonnet-4-6"
]
}耗时 1.8s · 严重度 info
通过PDF 文档识别PDF 内容识别正确权重 5 · 100›
PDF 内容识别正确
{
"status": 200,
"secret": "KIWI-1152",
"reply": "KIWI-1152"
}耗时 1.9s · 严重度 info
通过Prompt Caching缓存写入 3228 tokens、第二次命中 3228 tokens,Prompt Caching 完整可用权重 8 · 100›
缓存写入 3228 tokens、第二次命中 3228 tokens,Prompt Caching 完整可用
{
"first": {
"input": 9,
"creation": 3228,
"read": 0
},
"second": {
"input": 9,
"creation": 0,
"read": 3228
}
}耗时 4.7s · 严重度 info
通过长上下文召回长上下文召回全部通过(最高 32k)权重 10 · 100›
长上下文召回全部通过(最高 32k)
{
"tiers": [
{
"tokens": 32000,
"ok": true,
"status": 200,
"latencyMs": 3708,
"inputTokens": 3,
"reply": "728685",
"stop": "end_turn"
}
]
}耗时 3.7s · 严重度 info
这份报告对你有帮助吗?
评分 = Σ(单项分 × 权重) / Σ权重(跳过项不计)。≥70 通过,50–69 存在风险,<50 未达标;出现任一严重问题时结论封顶为「存在风险」。单次检测受网络与上游波动影响,建议结合该域名的历史记录判断。