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

2026/06/25 运行日期

claude-haiku-4-5-20251001 官方基准

1 通道分数 < 70 另含 1 任务失败

已揭晓揭晓进度

步骤 / 指标

PackyCode

cc-Zeus

官方基准

ClaudeZZ

R-kiro

OmniaKey

O-Max

AIMZ

M-Max

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 15:44:52 ~ 15:45:07 (10.1s) 15:45:00 ~ 15:45:05 (3.7s) 15:44:58 ~ 15:45:08 (6.1s) 15:44:54 ~ 15:45:04 (5.6s) 15:44:56 ~ 15:45:06 (6.5s)

提示词

查看本步 Prompt

ping. Reply with the single word: pong

HTTP 头(TTFB) 7,636 ms 2,114 ms 4,018 ms 3,182 ms 3,114 ms

首文本(SSE delta) 7,675 ms 2,116 ms 4,022 ms 4,012 ms 3,974 ms

总耗时 8,973 ms 2,396 ms 4,741 ms 4,079 ms 4,708 ms

输入 10 10 10 10 10

输出 46 49 58 52 51

缓存创建

33,121

5m: 0 / 1h: 33,121

33,122

5m: 0 / 1h: 33,122

22,404

5m: 0 / 1h: 22,404

33,192

5m: 0 / 1h: 33,192

32,785

5m: 0 / 1h: 32,785

缓存读取 0 0 0 0 0

输出速度 35.4 tok/s 175.0 tok/s 80.7 tok/s - 69.5 tok/s

响应

pong

pong

pong

pong

pong

步骤 2: identity text

时间 15:47:03 ~ 15:47:13 (5.2s) 15:46:50 ~ 15:47:00 (5.8s) 15:46:48 ~ 15:46:58 (6.8s) 15:48:29 ~ 15:48:39 (8.2s) 15:48:45 ~ 15:48:55 (7.8s)

提示词

查看本步 Prompt

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。自报身份，输出严格遵守以下 3 行格式（不允许 markdown、代码块、额外文本）：
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) 2,295 ms 1,872 ms 3,641 ms 2,751 ms 2,104 ms

首文本(SSE delta) 2,296 ms 1,874 ms 3,643 ms 4,188 ms 5,704 ms

总耗时 4,226 ms 4,449 ms 5,648 ms 6,881 ms 6,623 ms

输入 10 10 10 10 10

输出 278 282 205 254 421

缓存创建

240

5m: 0 / 1h: 240

243

5m: 0 / 1h: 243

259

5m: 0 / 1h: 259

246

5m: 0 / 1h: 246

245

5m: 0 / 1h: 245

缓存读取 33,121 33,122 22,404 33,192 32,785

输出速度 144.0 tok/s 109.5 tok/s 102.2 tok/s 94.3 tok/s burst

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

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

vendor: anthropic
brand: claude
model: claude-3-5-sonnet-20241022

步骤 3: cutoff text

时间 15:48:19 ~ 15:48:25 (3.2s) 15:50:46 ~ 15:50:51 (3.2s) 15:50:48 ~ 15:50:58 (6.7s) 15:51:34 ~ 15:51:44 (5.1s) 15:51:57 ~ 15:52:08 (5.6s)

提示词

查看本步 Prompt

Knowledge cutoff date? Reply with YYYY-MM only.

HTTP 头(TTFB) 1,624 ms 1,695 ms 4,305 ms 2,894 ms 2,388 ms

首文本(SSE delta) 1,625 ms 1,698 ms 4,308 ms 3,884 ms 2,718 ms

总耗时 2,282 ms 2,228 ms 5,064 ms 3,948 ms 3,557 ms

输入 10 10 10 10 10

输出 61 80 65 62 59

缓存创建

301

5m: 0 / 1h: 301

324

5m: 0 / 1h: 324

252

5m: 0 / 1h: 252

277

5m: 0 / 1h: 277

463

5m: 0 / 1h: 463

缓存读取 33,361 33,365 22,663 33,438 33,030

输出速度 92.8 tok/s 150.9 tok/s 86.0 tok/s - 70.3 tok/s

响应

2025-02

2025-02

2025-02

2025-02

2024-04

步骤 4: identity_free text

时间 15:49:49 ~ 15:50:14 (20.2s) 15:52:15 ~ 15:52:25 (7.1s) 15:53:03 ~ 15:53:18 (10.8s) 15:53:00 ~ 15:53:15 (13.7s) 15:53:43 ~ 15:53:53 (5.9s)

提示词

查看本步 Prompt

用中文回答：你是谁？由谁开发？50 字以内。

HTTP 头(TTFB) 2,081 ms 1,845 ms 4,563 ms 2,536 ms 2,242 ms

首文本(SSE delta) 2,082 ms 1,847 ms 4,564 ms 3,491 ms 3,394 ms

总耗时 19,176 ms 6,065 ms 9,775 ms 12,653 ms 3,671 ms

输入 10 10 10 10 10

输出 1,049 503 633 790 170

缓存创建

33,758

5m: 0 / 1h: 33,758

122

5m: 0 / 1h: 122

117

5m: 0 / 1h: 117

100

5m: 0 / 1h: 100

5m: 0 / 1h: 97

缓存读取 0 33,689 22,915 33,715 33,493

输出速度 61.4 tok/s 119.3 tok/s 121.5 tok/s 86.2 tok/s burst

响应

我是Claude Haiku 4.5，由Anthropic公司开发的AI助手。我可以帮助进行软件工程任务、代码分析和问题解答。知识截止日期为2025年2月。

我是Claude，由Anthropic开发的AI助手，擅长对话、编程、分析和解答问题。

我是Claude Code，由Anthropic开发的AI编程助手，帮助进行软件开发和代码调试。

我是Anthropic公司打造的Claude Haiku 4.5大型语言模型。我可以帮助进行编程、分析、文本处理等任务。知识库更新至2025年2月。

我是Claude，由Anthropic公司开发的AI助手。

步骤 5: knowledge_recall text

时间 15:53:29 ~ 15:54:14 (41.9s) 15:54:00 ~ 15:54:20 (15.8s) 15:55:28 ~ 15:55:43 (13.6s) 15:55:44 ~ 15:56:54 (65.5s) 15:57:48 ~ 15:58:13 (22.3s)

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) 5,405 ms 1,328 ms 4,187 ms 2,026 ms 1,825 ms

首文本(SSE delta) 5,475 ms 1,329 ms 4,189 ms 2,944 ms 19,105 ms

总耗时 40,466 ms 14,658 ms 12,786 ms 64,443 ms 19,959 ms

输入 10 10 10 10 10

输出 2,885 2,095 1,281 4,626 2,597

缓存创建

9,425

5m: 0 / 1h: 9,425

8,932

5m: 0 / 1h: 8,932

9,141

5m: 0 / 1h: 9,141

9,153

5m: 0 / 1h: 9,153

8,378

5m: 0 / 1h: 8,378

缓存读取 0 0 0 0 0

输出速度 82.4 tok/s 157.2 tok/s 149.0 tok/s 75.2 tok/s burst

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

{"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":"UNKNOWN","nba_2024":"UNKNOWN","copa_america_2024":"UNKNOWN","euro2024":"UNKNOWN","nobel_peace_2024

步骤 6: digit_count text

时间 15:56:27 ~ 15:59:48 (195.7s) 15:58:11 ~ 15:58:36 (22.3s) 15:58:20 ~ 15:58:40 (16.1s) 15:59:57 ~ 16:00:22 (22.2s) 16:01:30 ~ 16:01:45 (14.8s)

提示词

查看本步 Prompt

Count digits exactly in s. Return {"n":len(s),"c":[count0,count7,count9]}. s=820377914112075963475940468469322637064764775166681088801837391262783136

HTTP 头(TTFB) - 1,191 ms 2,451 ms 1,546 ms 1,175 ms

首文本(SSE delta) - 1,198 ms 2,453 ms 2,234 ms 1,176 ms

总耗时 - 21,266 ms 14,853 ms 20,363 ms 13,651 ms

输入 - 193 10 193 139

输出 - 3,777 1,751 3,795 2,284

缓存创建

5m: 0 / 1h: 0

5,980

5m: 0 / 1h: 5,980

5m: 0 / 1h: 0

缓存读取 - 0 0 0 0

输出速度 - 188.2 tok/s 141.2 tok/s 209.3 tok/s 183.1 tok/s

响应

失败

评测失败

```json
{"n":72,"c":[6,10,5]}
```

I'll count the digits in the string exactly.

```
s = "820377914112075963475940468469322637064764775166681088801837391262783136"
```

Counting:
- **n** (total length): 72 digits
- **c[0]** (count of 0

```json
{"n": 72, "c": [6, 10, 5]}
```

```json
{"n":72,"c":[6,10,5]}
```

总计

总输入(含缓存)

143,377

仅前 6 步

143,162

106,195

143,556

141,465

总输出

4,319

仅前 6 步

6,786

3,993

9,579

5,582

总缓存创建

76,845

5m: 0 / 1h: 76,845

仅前 6 步

42,743

5m: 0 / 1h: 42,743

38,153

5m: 0 / 1h: 38,153

42,968

5m: 0 / 1h: 42,968

41,968

5m: 0 / 1h: 41,968

总缓存读取

66,482

仅前 6 步

100,176

67,982

100,345

99,308

输出速度

77.2 tok/s

仅前 6 步

165.5 tok/s

134.5 tok/s

103.5 tok/s

170.4 tok/s

总执行时间

4m 36.3s

仅前 6 步

57.9s

1m 0.0s

2m 0.3s

1m 3.0s

总墙钟时间

14m 55.5s

含步骤间等待

仅前 6 步

13m 36.0s

含步骤间等待

13m 42.0s

含步骤间等待

15m 28.2s

含步骤间等待

16m 49.3s

含步骤间等待

按官方价目重估

同等内容若直发 Anthropic 的估算 ⓘ

$0.1243

Anthropic: Claude Haiku 4.5 (部分)

$0.0976

Anthropic: Claude Haiku 4.5

$0.0745

Anthropic: Claude Haiku 4.5

$0.1119

Anthropic: Claude Haiku 4.5

$0.0905

Anthropic: Claude Haiku 4.5

通道指纹

来源步骤 knowledge_recall digit_count digit_count digit_count digit_count

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

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

CDN cloudflare - - cloudflare cloudflare

ID 格式 msg_01CA5KgYXLePYVbPkbmUjd2c msg_01N9VMjGqdMkJHm3td1ukzBX msg_012v3cmo52zk3gV3EGdE4QGt msg_01PSQA1Fi5R9z2CLMhr69ACw msg_016zx3QM3b2asAiEPxxA316H

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-Zeus	via: 1.1 Caddy cf-ray: a112662039cc13a3-LAX server: cloudflare x-oneapi-request-id: 20260625155331219712061KnqXDh6A	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:x-oneapi-request-idhdr:server:cloudflarehdr:cf-rayhdr:viabody:id:msg_prefixbody:usage.service_tierbody:usage.inference_geobody:model:date_suffixbody:usage.cache_fieldsbody:extra_fieldsupstream:anthropic-direct-from-body	a112662039cc13a3-LAX
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_011CcPd5TJPYV9rzVz9p385Y
R-kiro	via: 1.1 google cf-ray: cc62e54356ef5043d5a3be3c14114db8-IAD server: nginx	id_format: msg_prefix via_chain: ["1.1 google"] 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	req_cc62e54356ef5043d5a3be3c14114db8 cc62e54356ef5043d5a3be3c14114db8-IAD
O-Max	cf-ray: a1126f9d78f2931d-AMS server: cloudflare x-oneapi-request-id: 20260625080000718077448268d9d60uCPi4BN	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-idhdr:server:cloudflarehdr:cf-raybody:id:msg_prefixbody:usage.service_tierbody:usage.inference_geobody:model:date_suffixbody:usage.cache_fieldsbody:extra_fieldsupstream:anthropic-direct-from-body	a1126f9d78f2931d-AMS
M-Max	cf-ray: a11271deda79fea2-AMS server: cloudflare x-new-api-version: v0.0.0 x-oneapi-request-id: 20260625080132228878838268d9d6HdrzYqQ9	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-new-api-versionhdr:x-oneapi-request-idhdr:server:cloudflarehdr:cf-raybody:id:msg_prefixbody:usage.service_tierbody:usage.inference_geobody:model:date_suffixbody:usage.cache_fieldsbody:extra_fieldsupstream:anthropic-direct-from-body	a11271deda79fea2-AMS

协议指纹评分（机器自动）

总分 / 100 100 92 97 77

基线状态已对照基线已对照基线已对照基线已对照基线

模型匹配 — 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

sliding 5m — 10.0 /10 ×13 10.0 /10 ×13 10.0 /10 ×13 10.0 /10 ×13

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

缓存命中比 — 10.0 /10 ×20 10.0 /10 ×20 10.0 /10 ×20 10.0 /10 ×20

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

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

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

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

Req-ID 透传 — 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

service_tier — 10.0 /10 ×6 10.0 /10 ×6 10.0 /10 ×6 10.0 /10 ×6

inference_geo — 10.0 /10 ×5 10.0 /10 ×5 10.0 /10 ×5 10.0 /10 ×5

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

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

延迟基线 — 10.0 /10 ×5 4.4 /10 ×5 9.0 /10 ×5 9.4 /10 ×5

流式投递 — 未计入未计入未计入未计入

综合结论

与基线相似度执行失败，不评分基准（参考） 92 v3.24.1 97 v3.24.1 77 v3.24.1

雷达图例

共 23 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 中文遵循 instruction_following_lang
22 档位思考量 tier_thinking_volume_match
23 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 15025 ms 8510 ms 8811 ms 18728 ms 8695 ms

输出速度 77.2 tok/s 165.5 tok/s 134.5 tok/s 103.5 tok/s 170.4 tok/s

完成步骤 5/6 成功 · 1 失败 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 ⓘ

$0.1243 (部分)

$0.0976

$0.0745

$0.1119

$0.0905

缓存读取占比 ⓘ 46% (部分) 70% 64% 70% 70%

揭晓

通道

服务商 PackyCode ↗

通道cc-Zeus

官方基准

服务商 ClaudeZZ ↗

通道R-kiro

服务商 OmniaKey ↗

通道O-Max

服务商 AIMZ ↗

通道M-Max

返回列表