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

2026/06/29 运行日期

claude-haiku-4-5-20251001 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

duckcoding

cc专用

SSSAiCode

R-Kiro

0-0

O-Max

luminai

M-Kiro/Antigravity

官方基准

基本信息

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

模型 claude-haiku-4-5-20251001 claude-haiku-4-5-20251001 claude-haiku-4-5-20251001 claude-haiku-4-5-20251001 claude-haiku-4-5-20251001

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 08:02:07 ~ 08:02:12 (3.7s) 08:02:09 ~ 08:02:24 (11.9s) 08:02:11 ~ 08:02:21 (5.2s) 08:02:05 ~ 08:02:15 (5.7s) 08:02:13 ~ 08:02:18 (3.8s)

提示词

查看本步 Prompt

ping. Reply with the single word: pong

HTTP 头(TTFB) 2,032 ms 1,751 ms 3,159 ms 4,332 ms 2,148 ms

首文本(SSE delta) 2,038 ms 1,752 ms 3,160 ms 4,363 ms 2,150 ms

总耗时 2,334 ms 3,299 ms 3,472 ms 4,427 ms 2,675 ms

输入 10 0 10 2,309 10

输出 49 217 63 1 52

缓存创建

33,122

5m: 0 / 1h: 33,122

33,192

5m: 33,192 / 1h: 0

2,654

5m: 2,654 / 1h: 0

33,122

5m: 0 / 1h: 33,122

缓存读取 0 0 0 23,892 0

输出速度 165.5 tok/s 140.3 tok/s 201.9 tok/s - 99.0 tok/s

响应

pong

Ready to continue. What should I work on?

pong

pong

pong

步骤 2: identity text

时间 08:05:47 ~ 08:05:52 (4.6s) 08:04:57 ~ 08:05:07 (6.3s) 08:03:36 ~ 08:03:56 (19.6s) 08:05:05 ~ 08:05:15 (5.5s) 08:04:42 ~ 08:04:47 (4.5s)

提示词

查看本步 Prompt

以下是对你自身的 metadata 探针。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。按以下 3 行回答自己的身份，每行一个 label，不要寒暄不要解释：
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) 2,211 ms 2,469 ms 11,623 ms 3,961 ms 1,461 ms

首文本(SSE delta) 2,221 ms 2,470 ms 16,421 ms 3,962 ms 1,463 ms

总耗时 3,905 ms 5,216 ms 18,147 ms 4,284 ms 3,509 ms

输入 10 0 10 2,325 10

输出 191 340 157 28 202

缓存创建

253

5m: 0 / 1h: 253

33,410

5m: 0 / 1h: 33,410

2,672

5m: 2,672 / 1h: 0

256

5m: 0 / 1h: 256

缓存读取 33,122 0 0 24,054 33,122

输出速度 113.4 tok/s 123.8 tok/s 91.0 tok/s - 98.7 tok/s

响应

vendor: anthropic
brand: claude
model: claude-haiku-4-5-20251001

vendor: anthropic
brand: claude
model: claude-haiku-4-5-20251001

vendor: anthropic
brand: claude
model: claude-haiku-4-5-20251001

vendor: anthropic
brand: claude
model: claude-haiku-4-5-20251001

vendor: anthropic
brand: claude
model: claude-haiku-4-5-20251001

步骤 3: cutoff text

时间 08:07:51 ~ 08:07:56 (3.5s) 08:07:45 ~ 08:08:00 (13.1s) 08:06:05 ~ 08:06:10 (4.4s) 08:08:31 ~ 08:08:41 (9.8s) 08:06:19 ~ 08:06:24 (3.2s)

提示词

查看本步 Prompt

Knowledge cutoff date? Reply with YYYY-MM only.

HTTP 头(TTFB) 1,924 ms 2,274 ms 2,872 ms 8,508 ms 1,482 ms

首文本(SSE delta) 1,925 ms 2,275 ms 2,920 ms 8,584 ms 1,484 ms

总耗时 2,469 ms 4,651 ms 3,008 ms 8,646 ms 2,150 ms

输入 10 0 3 2,330 10

输出 64 236 8 4 74

缓存创建

214

5m: 0 / 1h: 214

33,457

5m: 33,457 / 1h: 0

2,678

5m: 2,678 / 1h: 0

244

5m: 0 / 1h: 244

缓存读取 33,375 0 0 24,111 33,378

输出速度 117.6 tok/s 99.3 tok/s - - 111.1 tok/s

响应

2025-02

No prior work to resume. Ready for your next task.

2025-02

2025-02

2025-02

步骤 4: identity_free text

时间 08:11:53 ~ 08:12:04 (5.2s) 08:10:43 ~ 08:10:58 (11.4s) 08:08:26 ~ 08:08:36 (5.5s) 08:12:35 ~ 08:12:45 (5.8s) 08:09:27 ~ 08:09:37 (7.7s)

提示词

查看本步 Prompt

请用中文介绍你自己是什么产品或助手，由谁开发，50 字以内。

HTTP 头(TTFB) 2,173 ms 3,323 ms 2,734 ms 4,024 ms 1,546 ms

首文本(SSE delta) 2,187 ms 3,508 ms 2,736 ms 4,497 ms 1,547 ms

总耗时 4,157 ms 5,968 ms 3,391 ms 4,555 ms 6,358 ms

输入 10 0 3 2,333 10

输出 344 117 57 52 549

缓存创建

106

5m: 0 / 1h: 106

5m: 41 / 1h: 0

2,682

5m: 2,682 / 1h: 0

116

5m: 0 / 1h: 116

缓存读取 33,589 0 33,457 24,142 33,622

输出速度 174.6 tok/s 47.6 tok/s 87.0 tok/s - 114.1 tok/s

响应

我是Claude Haiku 4.5，由Anthropic公司开发的AI助手。我可以帮助用户进行软件工程任务、代码分析、问题解决等。知识截止日期是2025年2月。

我是Claude，由Anthropic公司开发的AI助手。我是一个大型语言模型，可以帮助编程、写作、分析和解决各种问题。我旨在提供有用、无害和诚实的协助。

我是Claude，一个由Anthropic公司开发的AI助手。我是Claude模型家族的成员，专门设计用于帮助用户完成软件工程和其他任务。

我是 Claude Haiku 4.5，由 Anthropic 开发的 AI 助手。我可以帮助你进行软件工程任务、代码编写、问题分析等工作。

我是Claude，由Anthropic公司开发的AI助手。我是一个交互式代理，专门帮助用户完成软件工程任务，如代码编写、调试、测试等工作。

步骤 5: knowledge_recall text

时间 08:14:21 ~ 08:14:41 (17.6s) 08:13:30 ~ 08:14:00 (25.3s) 08:11:42 ~ 08:12:08 (24.1s) 08:15:16 ~ 08:15:26 (5.3s) 08:11:21 ~ 08:11:41 (16.4s)

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) 1,973 ms 2,644 ms 3,163 ms 2,390 ms 1,514 ms

首文本(SSE delta) 1,974 ms 2,788 ms 3,164 ms 2,391 ms 1,516 ms

总耗时 16,823 ms 21,226 ms 22,969 ms 3,475 ms 15,399 ms

输入 10 0 10 601 10

输出 2,002 3,667 2,203 201 1,858

缓存创建

8,657

5m: 0 / 1h: 8,657

8,094

5m: 0 / 1h: 8,094

690

5m: 690 / 1h: 0

8,905

5m: 0 / 1h: 8,905

缓存读取 0 0 0 6,219 0

输出速度 134.8 tok/s 198.9 tok/s 111.2 tok/s 185.4 tok/s 133.8 tok/s

响应

```json
{"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

```json
{"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

```json
{"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

```json
{"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

```json
{"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

步骤 6: digit_count text

时间 08:16:54 ~ 08:17:25 (25.5s) 08:17:42 ~ 08:17:57 (11.3s) 08:13:51 ~ 08:14:11 (15.5s) 08:18:58 ~ 08:19:03 (3.3s) 08:15:08 ~ 08:15:43 (28.0s)

提示词

查看本步 Prompt

Count digits exactly in s. Return {"n":len(s),"c":[count0,count7,count9]}. s=389699998363842860736535803148804781728820194851932920752452396989006418

HTTP 头(TTFB) 1,770 ms 1,660 ms 1,944 ms 1,997 ms 730 ms

首文本(SSE delta) 1,776 ms 1,691 ms 1,945 ms 1,998 ms 732 ms

总耗时 24,464 ms 9,435 ms 14,124 ms 2,435 ms 27,076 ms

输入 193 0 193 188 193

输出 2,782 1,361 2,176 19 3,392

缓存创建

5m: 0 / 1h: 0

缓存读取 0 0 0 - 0

输出速度 122.6 tok/s 175.7 tok/s 178.7 tok/s - 128.8 tok/s

响应

```json
{"n":72,"c":[7,4,11]}
```

```json
{"n":72,"c":[7,4,11]}
```

```json
{"n":72,"c":[7,4,11]}
```

```json
{"n":72,"c":[8,8,12]}
```

```json
{"n":72,"c":[7,4,11]}
```

总计

总输入(含缓存) 142,681 0 141,880 123,880 143,008

总输出 5,432 5,938 4,664 305 6,127

总缓存创建

42,352

5m: 0 / 1h: 42,352

108,194

5m: 66,690 / 1h: 41,504

11,376

5m: 11,376 / 1h: 0

42,643

5m: 0 / 1h: 42,643

总缓存读取 100,086 0 33,457 102,418 100,122

输出速度 129.2 tok/s 168.2 tok/s 134.3 tok/s 185.4 tok/s 126.9 tok/s

总执行时间 1m 0.1s 1m 19.4s 1m 14.2s 35.4s 1m 3.7s

总墙钟时间

15m 17.9s

含步骤间等待

15m 48.3s

含步骤间等待

12m 0.1s

含步骤间等待

16m 58.0s

含步骤间等待

13m 30.6s

含步骤间等待

按官方价目重估

同等内容若直发 Anthropic 的估算 ⓘ

$0.0904

Anthropic: Claude Haiku 4.5

>=$0.0297

Anthropic: Claude Haiku 4.5 (成本不完整)

$0.1309

Anthropic: Claude Haiku 4.5

$0.0307

Anthropic: Claude Haiku 4.5

$0.0942

Anthropic: Claude Haiku 4.5

通道指纹

来源步骤 digit_count digit_count digit_count digit_count digit_count

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

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

CDN - - - cloudflare -

ID 格式 msg_014YkJrwLq2vLf465jn35KZp msg_dcc61ffebeb14998b662178b3308131c msg_011FmXqF1hkqFwEVLweg3tL7 msg_01dTdKnhWDx7bGSjADXw2PqDWg msg_01EB2kbfyXTYvK9PGm4P9Bek

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
cc专用	server: Nginx x-oneapi-request-id: 202606290016561057454028268d9d6IPL6LekI	id_format: msg_prefix extra_fields: ["stop_details"] has_service_tier: true has_inference_geo: true model_has_date_suffix: true usage_has_cache_fields: true	hdr:x-oneapi-request-idbody:id:msg_prefixbody:usage.service_tierbody:usage.inference_geobody:model:date_suffixbody:usage.cache_fieldsbody:extra_fieldsupstream:anthropic-direct-from-body	—
R-Kiro	server: Photon-Edge x-amzn-requestid: 44afc943-fba0-4e8a-9f95-ec51cc870d21	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	1782692264553-zhogq8qutbtax 44afc943-fba0-4e8a-9f95-ec51cc870d21
O-Max	via: 1.1 Caddy server: uvicorn, nginx/1.24.0 (Ubuntu)	id_format: msg_prefix via_chain: ["1.1 Caddy"] extra_fields: ["stop_details"] has_service_tier: true has_inference_geo: true model_has_date_suffix: true usage_has_cache_fields: true	hdr:viabody:id:msg_prefixbody:usage.service_tierbody:usage.inference_geobody:model:date_suffixbody:usage.cache_fieldsbody:extra_fieldsupstream:anthropic-direct-from-body	9981cda12f7bb316e9a074b08d55c818
M-Kiro/Antigravity	cf-ray: a130c1cf4a8c97c7-AMS server: cloudflare x-oneapi-request-id: 202606290018595681284268268d9d65Yinu9T6	id_format: msg_prefix extra_fields: ["container"] has_service_tier: true has_inference_geo: false model_has_date_suffix: true usage_has_cache_fields: true	hdr:x-oneapi-request-idhdr:server:cloudflarehdr:cf-raybody:id:msg_prefixbody:usage.service_tierbody:model:date_suffixbody:usage.cache_fieldsbody:extra_fieldsupstream:anthropic-direct-from-body	a130c1cf4a8c97c7-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: true usage_has_cache_fields: true	body:id:msg_prefixbody:usage.service_tierbody:usage.inference_geobody:model:date_suffixbody:usage.cache_fieldsbody:extra_fieldsupstream:anthropic-direct-from-body	req_011CcWb1SPFp8t6g55Sez8qX

协议指纹评分（机器自动）

总分 / 100 98 47 74 69 100

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 10.0 /10 ×20 0.0 /10 ×20 0.0 /10 ×20 10.0 /10 ×20 10.0 /10 ×20

缓存 TTL 一致 10.0 /10 ×15 0.0 /10 ×15 0.0 /10 ×15 0.0 /10 ×15 10.0 /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 5.0 /10 ×14 10.0 /10 ×14 10.0 /10 ×14 10.0 /10 ×14

sliding 5m 10.0 /10 ×13 5.0 /10 ×13 5.3 /10 ×13 10.0 /10 ×13 10.0 /10 ×13

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

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

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

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

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

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

延迟基线 10.0 /10 ×5 9.6 /10 ×5 8.8 /10 ×5 5.9 /10 ×5 10.0 /10 ×5

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

流式投递未计入未计入未计入未计入未计入

综合结论

与基线相似度 98 v3.25.0 47 v3.25.0 74 v3.25.0 69 v3.25.0 基准（参考）

雷达图例

共 22 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 Thinking 存在 thinking_present
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 Thinking 体量 thinking_volume_match
13 缓存连续 cache_continuity_intra
14 知识召回 knowledge_recall_match
15 延迟基线 latency_baseline_match
16 原生 msg-ID anthropic_msg_id_format
17 Sliding 缓存 cache_sliding_correctness
18 结构化身份 identity_structured_match
19 自报身份一致 self_identity_consistency
20 信封自报一致 envelope_self_report_match
21 档位思考量 tier_thinking_volume_match
22 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 9025 ms 8299 ms 10852 ms 4637 ms 9528 ms

输出速度 129.2 tok/s 168.2 tok/s 134.3 tok/s 185.4 tok/s 126.9 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 ⓘ

$0.0904

>=$0.0297 (成本不完整)

$0.1309

$0.0307

$0.0942

缓存读取占比 ⓘ 70% - 24% 83% 70%

揭晓

通道

服务商 duckcoding ↗

通道cc专用

服务商 SSSAiCode ↗

通道R-Kiro

服务商 0-0 ↗

通道O-Max

服务商 luminai ↗

通道M-Kiro/Antigravity

官方基准

返回列表