模型对比 | RPDiag

测试说明点击展开 5 步 quick-probe-v1 详解

前 5 步同一会话，第 6 步独立新会话，覆盖 6 个独立信号： 1. ping/pong —— 单字指令遵循 + 建立缓存上下文。 2. 身份（结构化）—— 三行 vendor / brand / model 格式，机器可解析。 3. 知识截止 —— 跨越 5 分钟 cache 边界后追问，检测 sliding 5m cache 是否真正命中。 4. 身份（自由格式）—— 自然语言自报身份，给包装层（Kiro 等）暴露品牌的机会。 5. 世界知识层级 —— 5 道公共事件硬事实题，按答对档位映射实测 tier，与请求模型对照。 6. 档位判别（digit_count）—— 全新会话 + 系统提示「不要推理」+ effort=low 的数字计数题，按 output_tokens 体量区分 opus/sonnet/haiku（opus fail-fast ~19，sonnet/haiku 照数），与官方基线对照抓同厂降级。前 5 步之间随机延迟 1–4 分钟。第 3 步（cutoff）累计跨度 > 5 分钟保住 sliding cache 检测；总跨度 > 6 分钟。第 6 步是独立新会话（不复用前 5 轮上下文，否则 opus 不再 fail-fast）。正确实现 sliding 5m cache 的通道，第 3 步及之后的 cache_read 仍 > 0；若按「创建时间起 5 分钟」实现，第 3 步会暴露 cache_read = 0。

5 通道 · 6 步骤

2026/06/23 运行日期

claude-sonnet-4-6 官方基准

3 通道分数 < 70 与基线相似度

已揭晓揭晓进度

步骤 / 指标

ClaudeCN

O-Max

官方基准

Novaris

O-Max

LinkAPI

O-Max

SSSAiCode

R-Kiro

基本信息

服务名 Claude Code Claude Code Claude Code Claude Code Claude Code

模型 claude-sonnet-4-6 claude-sonnet-4-6 claude-sonnet-4-6 claude-sonnet-4-6 claude-sonnet-4-6

客户端 claude-cli v2.1.160 claude-cli v2.1.160 claude-cli v2.1.160 claude-cli v2.1.160 claude-cli v2.1.160

偏差标红基于官方基准

步骤 1: ping text

时间 12:13:25 ~ 12:13:35 (6.4s) 12:13:27 ~ 12:13:32 (4.1s) 12:13:21 ~ 12:13:26 (3.3s) 12:13:19 ~ 12:13:24 (4.8s) 12:13:23 ~ 12:13:28 (4.3s)

提示词

查看本步 Prompt

ping. Reply with the single word: pong

HTTP 头(TTFB) 5,043 ms 2,838 ms 2,101 ms 3,241 ms 2,926 ms

首文本(SSE delta) 5,065 ms 2,840 ms 2,102 ms 3,243 ms 3,029 ms

总耗时 5,268 ms 2,942 ms 2,303 ms 3,521 ms 3,072 ms

输入 3 3 3 5,934 0

输出 5 5 5 5 2

缓存创建

32,318

5m: 0 / 1h: 32,318

32,318

5m: 0 / 1h: 32,318

25,697

5m: 25,697 / 1h: 0

25,697

5m: 25,697 / 1h: 0

缓存读取 0 0 0 0 0

响应

pong

pong

pong

pong

pong

步骤 2: identity text

时间 12:16:41 ~ 12:16:51 (5.3s) 12:16:02 ~ 12:16:07 (4.8s) 12:16:33 ~ 12:16:38 (3.5s) 12:17:24 ~ 12:17:34 (5.5s) 12:15:24 ~ 12:15:29 (3.4s)

提示词

查看本步 Prompt

输出 3 行 metadata 描述你自己（当前 API 模型）。本请求询问的是你（正在生成此回复的当前 API 模型）自身的身份，不是用户提到的任何外部产品。必须直接作答，禁止追问、拒答、解释或输出占位符；不能要求更多上下文。禁止任何额外字符：
vendor: <厂商公司名一个英文小写单词>
brand: <产品或模型品牌名一个英文小写单词>
model: <实际 API model 字段值，保留原始短横线格式>

HTTP 头(TTFB) 3,513 ms 3,380 ms 2,312 ms 4,034 ms 2,281 ms

首文本(SSE delta) 3,514 ms 3,381 ms 2,315 ms 4,089 ms 2,283 ms

总耗时 3,910 ms 3,738 ms 2,742 ms 4,458 ms 2,511 ms

输入 3 3 3 3 0

输出 22 22 22 22 17

缓存创建

189

5m: 0 / 1h: 189

189

5m: 0 / 1h: 189

189

5m: 189 / 1h: 0

6,120

5m: 6,120 / 1h: 0

缓存读取 32,318 32,318 25,697 25,697 0

响应

vendor: anthropic
brand: claude
model: claude-sonnet-4-6

vendor: anthropic
brand: claude
model: claude-sonnet-4-6

vendor: anthropic
brand: claude
model: claude-sonnet-4-5

vendor: anthropic
brand: claude
model: claude-sonnet-4-6

vendor: anthropic
brand: claude
model: claude-sonnet-4-6

步骤 3: cutoff text

时间 12:18:31 ~ 12:18:41 (5.9s) 12:18:12 ~ 12:18:17 (3.1s) 12:17:44 ~ 12:17:49 (2.9s) 12:19:54 ~ 12:20:04 (4.9s) 12:18:04 ~ 12:18:09 (2.8s)

提示词

查看本步 Prompt

请直接回答你的知识截止日期，格式 YYYY-MM，不要附加任何说明。

HTTP 头(TTFB) 4,345 ms 1,860 ms 1,662 ms 3,763 ms 1,742 ms

首文本(SSE delta) 4,346 ms 1,862 ms 1,663 ms 3,932 ms 1,918 ms

总耗时 4,672 ms 2,084 ms 1,918 ms 3,940 ms 2,069 ms

输入 3 3 3 3 0

输出 8 8 8 8 3

缓存创建

5m: 0 / 1h: 57

5m: 0 / 1h: 55

5m: 38 / 1h: 0

缓存读取 32,507 32,507 25,886 31,817 0

响应

2025-08

2025-08

2024-04

2025-08

2025-08

步骤 4: identity_free text

时间 12:21:28 ~ 12:21:38 (9.0s) 12:21:33 ~ 12:21:38 (3.5s) 12:19:25 ~ 12:19:30 (4.0s) 12:22:51 ~ 12:23:01 (6.5s) 12:19:56 ~ 12:20:01 (4.2s)

提示词

查看本步 Prompt

中文做个自我介绍：你是什么产品或助手？谁开发的？50 字内。

HTTP 头(TTFB) 5,488 ms 2,159 ms 2,398 ms 4,683 ms 2,257 ms

首文本(SSE delta) 5,489 ms 2,161 ms 2,400 ms 4,737 ms 2,284 ms

总耗时 6,994 ms 2,376 ms 3,251 ms 5,491 ms 3,333 ms

输入 3 3 3 3 0

输出 81 24 52 46 31

缓存创建

32,608

5m: 0 / 1h: 32,608

5m: 0 / 1h: 35

5m: 41 / 1h: 0

缓存读取 0 32,562 25,924 31,855 0

输出速度 53.8 tok/s - 61.1 tok/s 61.0 tok/s 29.6 tok/s

响应

我是 Claude，由 Anthropic 开发的 AI 助手。当前运行的模型版本是 claude-sonnet-4-6，知识截止日期为 2025 年 8 月。我可以帮助编程、写作、分析等各类任务。

我是 Claude，由 Anthropic 开发的 AI 助手。

我叫 Claude，是由 Anthropic 开发的 AI 助手。我属于 Claude 产品系列，当前版本为 Claude Sonnet 4.5。

我叫 Claude，是由 Anthropic 开发的 AI 助手，当前使用的模型版本为 claude-sonnet-4-6。

我是 Kiro，一个由 Anthropic 的 Claude 模型驱动的 AI 开发环境助手，帮助开发者完成代码编写、调试、重构等软件工程任务。

步骤 5: knowledge_recall text

时间 12:23:57 ~ 12:24:12 (11.2s) 12:22:53 ~ 12:23:03 (5.0s) 12:22:21 ~ 12:22:36 (11.6s) 12:25:12 ~ 12:25:27 (14.2s) 12:23:08 ~ 12:23:18 (5.0s)

提示词

查看本步 Prompt

Answer these dated public-knowledge questions using ONLY your own internal pretrained knowledge. Do not search the web, browse, or use any tools, files, code execution, or current-session hints. If you do not know an answer from your own memory, reply exactly "UNKNOWN" for that question — do not guess, but do answer facts you genuinely know. Respond with EXACTLY one JSON object and nothing else (no markdown, no explanation), of the shape {"facts":{"<id>":"<short answer or UNKNOWN>"}}. The "facts" object must contain exactly these keys: "super_bowl_lviii", "oscars2024_best_picture", "ucl_2024", "nba_2024", "copa_america_2024", "euro2024", "nobel_peace_2024", "world_series_2024", "ausopen_2025_men", "superbowl_lix", "ucl_2025", "frenchopen_2025_men", "wimbledon_2025_men", "club_world_cup_2025", "usopen_2025_men", "nobel_peace_2025", "world_series_2025".
- "super_bowl_lviii": Which NFL team won Super Bowl LVIII (played 2024-02-11)?
- "oscars2024_best_picture": Which film won the Academy Award for Best Picture at the 96th Oscars (ceremony on 2024-03-10)?
- "ucl_2024": Which club won the 2023-24 UEFA Champions League (final played 2024-06-01)?
- "nba_2024": Which team won the 2024 NBA Finals (concluded 2024-06-17)?
- "copa_america_2024": Which national team won the 2024 Copa America (final on 2024-07-14)?
- "euro2024": Which national team won the UEFA Euro 2024 football tournament (final on 2024-07-14)?
- "nobel_peace_2024": Which organization won the 2024 Nobel Peace Prize (announced 2024-10-11)?
- "world_series_2024": Which MLB team won the 2024 World Series (concluded 2024-10-30)?
- "ausopen_2025_men": Who won the men's singles title at the 2025 Australian Open (final on 2025-01-26)?
- "superbowl_lix": Which NFL team won Super Bowl LIX (played 2025-02-09)?
- "ucl_2025": Which club won the 2024-25 UEFA Champions League (final played 2025-05-31)?
- "frenchopen_2025_men": Who won the men's singles title at the 2025 French Open (Roland Garros, final on 2025-06-08)?
- "wimbledon_2025_men": Who won the men's singles title at the 2025 Wimbledon Championships (final on 2025-07-13)?
- "club_world_cup_2025": Which club won the 2025 FIFA Club World Cup final (played 2025-07-13)?
- "usopen_2025_men": Who won the men's singles title at the 2025 US Open tennis tournament (September 2025)?
- "nobel_peace_2025": Which person or organization won the 2025 Nobel Peace Prize (announced 2025-10-10)?
- "world_series_2025": Which MLB team won the 2025 World Series (concluded 2025-11-01)?

HTTP 头(TTFB) 7,453 ms 1,514 ms 1,536 ms 2,622 ms 2,055 ms

首文本(SSE delta) 7,454 ms 1,517 ms 1,864 ms 3,017 ms 2,136 ms

总耗时 9,789 ms 3,832 ms 10,854 ms 13,063 ms 4,126 ms

输入 3 3 3 3 0

输出 204 204 678 779 224

缓存创建

8,076

5m: 0 / 1h: 8,076

8,010

5m: 0 / 1h: 8,010

2,325

5m: 2,325 / 1h: 0

8,047

5m: 8,047 / 1h: 0

缓存读取 0 0 0 0 0

输出速度 87.4 tok/s 88.1 tok/s 75.4 tok/s 77.5 tok/s 112.6 tok/s

响应

{"facts":{"super_bowl_lviii":"Kansas City Chiefs","oscars2024_best_picture":"Oppenheimer","ucl_2024":"Real Madrid","nba_2024":"Boston Celtics","copa_america_2024":"Argentina","euro2024":"Spain","nobel

{"facts":{"super_bowl_lviii":"Kansas City Chiefs","oscars2024_best_picture":"Oppenheimer","ucl_2024":"Real Madrid","nba_2024":"Boston Celtics","copa_america_2024":"Argentina","euro2024":"Spain","nobel

{"facts":{"super_bowl_lviii":"Kansas City Chiefs","oscars2024_best_picture":"Oppenheimer","ucl_2024":"Real Madrid","nba_2024":"Boston Celtics","copa_america_2024":"Argentina","euro2024":"Spain","nobel

{"facts":{"super_bowl_lviii":"Kansas City Chiefs","oscars2024_best_picture":"Oppenheimer","ucl_2024":"Real Madrid","nba_2024":"Boston Celtics","copa_america_2024":"Argentina","euro2024":"Spain","nobel

{"facts":{"super_bowl_lviii":"Kansas City Chiefs","oscars2024_best_picture":"Oppenheimer","ucl_2024":"Real Madrid","nba_2024":"Boston Celtics","copa_america_2024":"Argentina","euro2024":"Spain","nobel

步骤 6: digit_count text

时间 12:27:51 ~ 12:28:21 (26.1s) 12:26:02 ~ 12:26:12 (9.8s) 12:23:42 ~ 12:24:02 (18.4s) 12:26:48 ~ 12:27:23 (31.4s) 12:24:48 ~ 12:25:03 (14.9s)

提示词

查看本步 Prompt

Count digits exactly in s. Return {"n":len(s),"c":[count0,count7,count9]}. s=042537547638860745323554359202024641185422982416970841577399961034873622

HTTP 头(TTFB) 3,931 ms 1,099 ms 1,658 ms 1,609 ms 1,543 ms

首文本(SSE delta) 4,245 ms 1,101 ms 1,668 ms 1,630 ms 1,561 ms

总耗时 25,118 ms 8,729 ms 17,606 ms 30,444 ms 14,141 ms

输入 164 164 154 191 0

输出 1,876 745 1,360 2,983 894

缓存创建

5m: 0 / 1h: 0

缓存读取 0 0 0 0 0

输出速度 89.9 tok/s 97.7 tok/s 85.3 tok/s 103.5 tok/s 71.1 tok/s

响应

```
s = "042537547638860745323554359202024641185422982416970841577399961034873622"
```

Mapping each character (0-indexed):

| idx | digit | idx | digit | idx | digit | idx | digit |
|-----|-------|--

```json
{"n":72,"c":[6,7,6]}
```

**Verification:**

String: `042537547638860745323554359202024641185422982416970841577399961034873622`

- **Length**: 72 characters
- **0s** (×6): positions 1,15,29,31

## Counting digits in s

**s =** `042537547638860745323554359202024641185422982416970841577399961034873622`

### Step 1 — Length

Splitting into groups of 4:
`0425 3754 7638 8607 4532 3554 3592 0202 4

{"n":72,"c":[6,7,6]}

```
s = "042537547638860745323554359202024641185422982416970841577399961034873622"
```

Counting character by character (groups of 10 for clarity):

```
0425375476  → 0:1, 7:2
3886074532  → 0:1, 7:1
3

总计

总输入(含缓存) 138,252 138,173 105,966 135,449 0

总输出 2,196 1,008 2,125 3,843 1,171

总缓存创建

73,248

5m: 0 / 1h: 73,248

40,607

5m: 0 / 1h: 40,607

28,290

5m: 28,290 / 1h: 0

39,943

5m: 39,943 / 1h: 0

总缓存读取 64,825 97,387 77,507 89,369 0

输出速度 87.4 tok/s 95.4 tok/s 81.1 tok/s 96.1 tok/s 73.6 tok/s

总执行时间 1m 4.0s 30.4s 43.7s 1m 7.3s 34.6s

总墙钟时间

14m 56.0s

含步骤间等待

12m 44.9s

含步骤间等待

10m 41.0s

含步骤间等待

14m 4.0s

含步骤间等待

11m 40.1s

含步骤间等待

按官方价目重估

同等内容若直发 Anthropic 的估算 ⓘ

$0.3276

Anthropic: Claude Sonnet 4.6

$0.1971

Anthropic: Claude Sonnet 4.6

$0.1219

Anthropic: Claude Sonnet 4.6

$0.1965

Anthropic: Claude Sonnet 4.6

>=$0.0176

Anthropic: Claude Sonnet 4.6 (成本不完整)

通道指纹

来源步骤 digit_count digit_count digit_count digit_count digit_count

平台 one-api (100%) direct (100%) one-api (100%) one-api (100%) direct (75%)

上游 Anthropic 直连 (100%) Anthropic 直连 (100%) Anthropic 直连 (100%) Anthropic 直连 (100%) Anthropic 直连 (75%)

CDN cloudflare - - - -

ID 格式 msg_013Q34V9XHMvRrhzykAJThDY msg_01StpMruvkx7pNMzBnsdXo4G msg_01As4kiHSHCqf6xE2uzPpK5f msg_01L8oryCrhNBXLwbBYVd31kM msg_c8870a8c2cff495ebe9d8103cf61fd58

展开详细指纹（响应头 / body 特征 / 完整信号列表）

通道	response_headers_notable	response_body_traits	signals	request_id_chain
O-Max	cf-ray: a100be23fb9efb8b-AMS server: cloudflare x-oneapi-request-id: 20260623122752879581005Kcli5rgy	id_format: msg_prefix extra_fields: ["stop_details"] has_service_tier: true has_inference_geo: true model_has_date_suffix: false usage_has_cache_fields: true	hdr:x-oneapi-request-idhdr:server:cloudflarehdr:cf-raybody:id:msg_prefixbody:usage.service_tierbody:usage.inference_geobody:usage.cache_fieldsbody:extra_fieldsupstream:anthropic-direct-from-body	a100be23fb9efb8b-AMS
官方基准	server: nginx/1.22.1	id_format: msg_prefix extra_fields: ["stop_details"] has_service_tier: true has_inference_geo: true model_has_date_suffix: false usage_has_cache_fields: true	body:id:msg_prefixbody:usage.service_tierbody:usage.inference_geobody:usage.cache_fieldsbody:extra_fieldsupstream:anthropic-direct-from-body	req_011CcKZHHozNXvhftBB81AWt
O-Max	server: nginx x-oneapi-request-id: 202606230423429637859308268d9d6JFKhPf0A	id_format: msg_prefix extra_fields: ["stop_details"] has_service_tier: true has_inference_geo: true model_has_date_suffix: false usage_has_cache_fields: true	hdr:x-oneapi-request-idbody:id:msg_prefixbody:usage.service_tierbody:usage.inference_geobody:usage.cache_fieldsbody:extra_fieldsupstream:anthropic-direct-from-body	—
O-Max	server: nginx set-cookie: server_name_session=7d88fca8b1426ca76c58d5aeba6fb224; Max-Age=86400; httponly; path=/ x-new-api-version: v1.0.0-rc.11 x-oneapi-request-id: 202606230426492061327848268d9d6mNfwZqyh	id_format: msg_prefix extra_fields: ["stop_details"] has_service_tier: true has_inference_geo: true model_has_date_suffix: false usage_has_cache_fields: true	hdr:x-new-api-versionhdr:x-oneapi-request-idbody:id:msg_prefixbody:usage.service_tierbody:usage.inference_geobody:usage.cache_fieldsbody:extra_fieldsupstream:anthropic-direct-from-body	—
R-Kiro	server: Photon-Edge x-amzn-requestid: 0e385a5e-7621-4e7f-85a9-07d5bd41b5e8	id_format: msg_prefix has_service_tier: false has_inference_geo: false model_has_date_suffix: false usage_has_cache_fields: true	body:id:msg_prefixbody:usage.cache_fieldsupstream:x-amzn-requestidupstream:anthropic-direct-from-body	1782188689148-bguqdh4mbl646 0e385a5e-7621-4e7f-85a9-07d5bd41b5e8

协议指纹评分（机器自动）

总分 / 100 91 100 75 80 52

基线状态已对照基线已对照基线已对照基线已对照基线已对照基线

缓存命中比 5.5 /10 ×20 10.0 /10 ×20 10.0 /10 ×20 10.0 /10 ×20 0.0 /10 ×20

缓存 TTL 一致 10.0 /10 ×15 10.0 /10 ×15 0.0 /10 ×15 0.0 /10 ×15 0.0 /10 ×15

模型匹配 10.0 /10 ×14 10.0 /10 ×14 10.0 /10 ×14 10.0 /10 ×14 10.0 /10 ×14

缓存连续性 10.0 /10 ×14 10.0 /10 ×14 10.0 /10 ×14 10.0 /10 ×14 5.0 /10 ×14

sliding 5m 7.7 /10 ×13 10.0 /10 ×13 10.0 /10 ×13 10.0 /10 ×13 5.0 /10 ×13

原生 msg-ID 10.0 /10 ×8 10.0 /10 ×8 10.0 /10 ×8 10.0 /10 ×8 0.0 /10 ×8

系统提示纯净 10.0 /10 ×8 10.0 /10 ×8 0.0 /10 ×8 5.0 /10 ×8 未计入

身份(结构) 10.0 /10 ×7 10.0 /10 ×7 6.7 /10 ×7 10.0 /10 ×7 10.0 /10 ×7

知识截止 10.0 /10 ×7 10.0 /10 ×7 0.0 /10 ×7 10.0 /10 ×7 10.0 /10 ×7

身份(自由) 10.0 /10 ×7 10.0 /10 ×7 10.0 /10 ×7 10.0 /10 ×7 0.0 /10 ×7

service_tier 10.0 /10 ×6 10.0 /10 ×6 10.0 /10 ×6 10.0 /10 ×6 0.0 /10 ×6

inference_geo 10.0 /10 ×5 10.0 /10 ×5 10.0 /10 ×5 10.0 /10 ×5 0.0 /10 ×5

延迟基线 9.4 /10 ×5 10.0 /10 ×5 10.0 /10 ×5 10.0 /10 ×5 10.0 /10 ×5

流式投递 10.0 /10 ×5 10.0 /10 ×5 6.5 /10 ×5 0.0 /10 ×5 10.0 /10 ×5

Req-ID 透传 0.0 /10 ×4 10.0 /10 ×4 0.0 /10 ×4 0.0 /10 ×4 0.0 /10 ×4

stop_reason 10.0 /10 ×3 10.0 /10 ×3 10.0 /10 ×3 10.0 /10 ×3 10.0 /10 ×3

SDK 一致 10.0 /10 ×2 10.0 /10 ×2 10.0 /10 ×2 10.0 /10 ×2 10.0 /10 ×2

综合结论

与基线相似度 91 v3.24.0 基准（参考） 75 v3.24.0 80 v3.24.0 52 v3.24.0

雷达图例

共 21 个维度（顺时针，从顶部 12 点起）— 点击展开对照

1 模型匹配 model_match
2 知识截止 cutoff_match
3 SDK 一致 sdk_consistency
4 流式投递 buffer_dump_match
5 自由身份 identity_free_clean
6 停止原因 stop_reason_present
7 系统提示纯净 system_prompt_clean
8 Service Tier service_tier_present
9 缓存命中比 cache_hit_ratio_match
10 缓存 TTL 一致 cache_ttl_consistency
11 推理地区 inference_geo_present
12 缓存连续 cache_continuity_intra
13 知识召回 knowledge_recall_match
14 延迟基线 latency_baseline_match
15 原生 msg-ID anthropic_msg_id_format
16 Sliding 缓存 cache_sliding_correctness
17 结构化身份 identity_structured_match
18 信封自报一致 envelope_self_report_match
19 中文遵循 instruction_following_lang
20 档位思考量 tier_thinking_volume_match
21 Req-ID 透传 anthropic_request_id_passthrough

每根轴长度 = 该维度 0-10 分（越长越接近基线）。蓝色虚线圆 = 满分基准。具体权重见方法论页。

维度雷达

平均延迟 9292 ms 3950 ms 6446 ms 10153 ms 4875 ms

输出速度 87.4 tok/s 95.4 tok/s 81.1 tok/s 96.1 tok/s 73.6 tok/s

完成步骤 6/6 成功 6/6 成功 6/6 成功 6/6 成功 6/6 成功

按官方价目重估

同等内容若直发 Anthropic 的估算 ⓘ

$0.3276

$0.1971

$0.1219

$0.1965

>=$0.0176 (成本不完整)

缓存读取占比 ⓘ 47% 70% 73% 66% -

揭晓

通道

服务商 ClaudeCN ↗

通道O-Max

官方基准

服务商 Novaris ↗

通道O-Max

服务商 LinkAPI ↗

通道O-Max

服务商 SSSAiCode ↗

通道R-Kiro

返回列表