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

2026/06/27 运行日期

claude-haiku-4-5-20251001 官方基准

1 通道分数 < 70 与基线相似度

已揭晓揭晓进度

步骤 / 指标

AIMZ

M-Max

LingxiCode

M-

OmniaKey

O-Max

duckcoding

cc专用

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 18:20:15 ~ 18:20:20 (4.9s) 18:20:09 ~ 18:20:19 (5.0s) 18:20:13 ~ 18:20:23 (8.2s) 18:20:11 ~ 18:20:21 (5.3s) 18:20:17 ~ 18:20:27 (5.1s)

提示词

查看本步 Prompt

ping. Reply with the single word: pong

HTTP 头(TTFB) 3,166 ms 3,051 ms 5,067 ms 4,046 ms 4,088 ms

首文本(SSE delta) 3,514 ms 3,055 ms 5,068 ms 4,212 ms 4,089 ms

总耗时 3,629 ms 3,553 ms 6,228 ms 4,594 ms 4,096 ms

输入 10 10 10 10 10

输出 57 52 49 59 48

缓存创建

32,785

5m: 0 / 1h: 32,785

33,124

5m: 0 / 1h: 33,124

33,122

5m: 0 / 1h: 33,122

33,122

5m: 0 / 1h: 33,122

33,122

5m: 0 / 1h: 33,122

缓存读取 0 0 0 0 0

输出速度 - 104.4 tok/s 42.2 tok/s 154.5 tok/s burst

响应

pong

pong

pong

pong

pong

步骤 2: identity text

时间 18:23:17 ~ 18:23:32 (12.1s) 18:22:36 ~ 18:22:46 (5.5s) 18:22:30 ~ 18:22:40 (7.0s) 18:23:19 ~ 18:23:29 (6.7s) 18:22:52 ~ 18:23:02 (6.9s)

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) 2,576 ms 2,249 ms 3,631 ms 3,656 ms 4,341 ms

首文本(SSE delta) 10,586 ms 2,252 ms 3,633 ms 3,858 ms 4,342 ms

总耗时 10,876 ms 4,285 ms 5,308 ms 5,973 ms 5,864 ms

输入 10 10 10 10 10

输出 606 203 162 286 190

缓存创建

248

5m: 0 / 1h: 248

243

5m: 0 / 1h: 243

240

5m: 0 / 1h: 240

250

5m: 0 / 1h: 250

239

5m: 0 / 1h: 239

缓存读取 32,785 33,124 33,122 33,122 33,122

输出速度 burst 99.9 tok/s 96.7 tok/s 135.2 tok/s 124.8 tok/s

响应

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

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 18:26:28 ~ 18:26:33 (4.6s) 18:25:25 ~ 18:25:30 (4.8s) 18:25:50 ~ 18:26:00 (6.2s) 18:24:38 ~ 18:25:03 (24.1s) 18:25:35 ~ 18:25:45 (6.1s)

提示词

查看本步 Prompt

Knowledge cutoff date? Reply with YYYY-MM only.

HTTP 头(TTFB) 2,783 ms 3,051 ms 3,685 ms 14,645 ms 4,361 ms

首文本(SSE delta) 3,190 ms 3,052 ms 3,686 ms 15,817 ms 4,363 ms

总耗时 3,444 ms 3,943 ms 4,540 ms 23,235 ms 5,138 ms

输入 10 10 10 10 10

输出 65 66 62 73 83

缓存创建

629

5m: 0 / 1h: 629

243

5m: 0 / 1h: 243

185

5m: 0 / 1h: 185

326

5m: 0 / 1h: 326

232

5m: 0 / 1h: 232

缓存读取 33,033 33,367 33,362 33,372 33,361

输出速度 - 74.1 tok/s 72.6 tok/s 9.8 tok/s 107.1 tok/s

响应

2024-04

2025-02

2025-02

2025-02

2025-02

步骤 4: identity_free text

时间 18:27:59 ~ 18:28:10 (8.2s) 18:28:30 ~ 18:28:40 (7.3s) 18:27:55 ~ 18:28:10 (13.3s) 18:27:17 ~ 18:27:37 (16.1s) 18:29:12 ~ 18:29:17 (4.7s)

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) 2,808 ms 2,999 ms 3,424 ms 3,383 ms 1,612 ms

首文本(SSE delta) 6,915 ms 3,000 ms 3,425 ms 3,570 ms 1,614 ms

总耗时 7,096 ms 6,469 ms 12,376 ms 15,246 ms 3,670 ms

输入 10 10 10 10 10

输出 712 444 1,321 373 233

缓存创建

107

5m: 0 / 1h: 107

104

5m: 0 / 1h: 104

100

5m: 0 / 1h: 100

111

5m: 0 / 1h: 111

33,716

5m: 0 / 1h: 33,716

缓存读取 33,662 33,610 33,547 33,698 0

输出速度 burst 128.0 tok/s 147.6 tok/s 31.9 tok/s 113.3 tok/s

响应

我是Claude，由Anthropic开发的AI助手。我可以帮助完成编程、写作、分析等任务。

我是 Claude Haiku 4.5，由 Anthropic 公司打造的 AI 助手。

我是Claude Haiku 4.5，由Anthropic公司打造的AI助手。专长于代码编辑、文本处理、数据分析等任务。

我是Claude Haiku 4.5，由Anthropic公司打造的AI助手。Haiku是Claude系列中最轻量级的模型，专门优化用于快速、高效的对话和编程协助。

我是Claude Haiku 4.5，属于Anthropic公司开发的Claude AI助手产品系列。Anthropic是一家AI安全公司，致力于开发可靠、可解释的AI系统。

步骤 5: knowledge_recall text

时间 18:31:52 ~ 18:32:07 (14.1s) 18:31:35 ~ 18:31:55 (19.4s) 18:29:36 ~ 18:30:11 (34.3s) 18:28:55 ~ 18:29:25 (27.1s) 18:32:31 ~ 18:32:56 (24.9s)

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) 3,208 ms 2,212 ms 4,068 ms 10,644 ms 1,857 ms

首文本(SSE delta) 11,241 ms 2,213 ms 4,069 ms 11,634 ms 1,859 ms

总耗时 11,936 ms 18,481 ms 32,091 ms 26,353 ms 23,893 ms

输入 10 10 10 10 10

输出 1,435 2,823 4,220 1,993 2,618

缓存创建

9,099

5m: 0 / 1h: 9,099

8,776

5m: 0 / 1h: 8,776

9,586

5m: 0 / 1h: 9,586

8,800

5m: 0 / 1h: 8,800

8,567

5m: 0 / 1h: 8,567

缓存读取 0 0 0 0 0

输出速度 burst 173.5 tok/s 150.6 tok/s 135.4 tok/s 118.8 tok/s

响应

```json
{"facts":{"super_bowl_lviii":"Kansas City Chiefs","oscars2024_best_picture":"Oppenheimer","ucl_2024":"UNKNOWN","nba_2024":"UNKNOWN","copa_america_2024":"UNKNOWN","euro2024":"UNKNOWN","nobel_pe

```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":"Real Madrid","nba_2024":"Boston Celtics","copa_america_2024":"Argentina","euro2024":"Spain","nobel

步骤 6: digit_count text

时间 18:33:14 ~ 18:33:29 (14.5s) 18:34:34 ~ 18:34:59 (20.6s) 18:32:17 ~ 18:32:52 (32.6s) 18:31:27 ~ 18:32:07 (35.0s) 18:34:07 ~ 18:34:32 (20.0s)

提示词

查看本步 Prompt

Count digits exactly in s. Return {"n":len(s),"c":[count0,count7,count9]}. s=434658761558407277976518041535396432400359335247033327167252071701525791

HTTP 头(TTFB) 1,851 ms 1,849 ms 2,242 ms 1,706 ms 1,215 ms

首文本(SSE delta) 1,851 ms 1,850 ms 2,243 ms 1,755 ms 1,218 ms

总耗时 13,451 ms 19,892 ms 31,574 ms 34,315 ms 18,991 ms

输入 139 193 193 193 193

输出 2,069 3,457 3,625 6,201 3,196

缓存创建

5m: 0 / 1h: 0

缓存读取 0 0 0 0 0

输出速度 178.4 tok/s 191.6 tok/s 123.6 tok/s 190.4 tok/s 179.8 tok/s

响应

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

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

```json
{"n":73,"c":[8,11,4]}
```

```json
{"n":73,"c":[8,11,4]}
```

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

总计

总输入(含缓存) 142,537 142,834 143,507 143,044 142,602

总输出 4,944 7,045 9,439 8,985 6,368

总缓存创建

42,868

5m: 0 / 1h: 42,868

42,490

5m: 0 / 1h: 42,490

43,233

5m: 0 / 1h: 43,233

42,609

5m: 0 / 1h: 42,609

75,876

5m: 0 / 1h: 75,876

总缓存读取 99,480 100,101 100,031 100,192 66,483

输出速度 178.4 tok/s 171.0 tok/s 134.9 tok/s 130.5 tok/s 143.1 tok/s

总执行时间 58.4s 1m 2.5s 1m 41.7s 1m 54.4s 1m 7.8s

总墙钟时间

13m 14.0s

含步骤间等待

14m 50.1s

含步骤间等待

12m 39.1s

含步骤间等待

11m 56.1s

含步骤间等待

14m 15.1s

含步骤间等待

按官方价目重估

同等内容若直发 Anthropic 的估算 ⓘ

$0.0884

Anthropic: Claude Haiku 4.5

$0.0986

Anthropic: Claude Haiku 4.5

$0.1115

Anthropic: Claude Haiku 4.5

$0.1084

Anthropic: Claude Haiku 4.5

$0.1336

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%) one-api (100%) one-api (100%) one-api (100%) direct (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_01ShZdiYu9khYJ9wAHuutMwu msg_01W4NBr971FSQee3WTyjHVQd msg_01Y3myEUiohTt2qzDEiuaya4 msg_01YFRZq6GK3bFMejR4qsnQRw msg_018qa9L4gWrQzMoAdG8Um1K7

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
M-Max	cf-ray: a123cadfccb9bccf-AMS server: cloudflare x-new-api-version: v0.0.0 x-oneapi-request-id: 202606271033155293628448268d9d68L7ME2mQ	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	a123cadfccb9bccf-AMS
M-	cf-ray: a123ccd1ff39e393-NRT server: cloudflare set-cookie: server_name_session=d962c7fc2e4cdc3f87113d06b76bd7cf; Max-Age=86400; httponly; path=/ x-oneapi-request-id: 20260627103435824077989OKhnLIbl	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	a123ccd1ff39e393-NRT
O-Max	cf-ray: a123c97c5ecb29c4-AMS server: cloudflare x-oneapi-request-id: 20260627103219934809548268d9d6tXLm3cZI	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	a123c97c5ecb29c4-AMS
cc专用	server: Nginx x-oneapi-request-id: 202606271031284613212188268d9d6qNojRos5	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	—
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_011CcTcawzWRiwFEi8mexedS

协议指纹评分（机器自动）

总分 / 100 80 97 97 97 93

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 10.0 /10 ×20 10.0 /10 ×20 10.0 /10 ×20 5.5 /10 ×20

缓存 TTL 一致 10.0 /10 ×15 10.0 /10 ×15 10.0 /10 ×15 10.0 /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 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 7.7 /10 ×13

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

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

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

知识截止 0.0 /10 ×7 10.0 /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 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 10.0 /10 ×5

延迟基线 9.4 /10 ×5 10.0 /10 ×5 10.0 /10 ×5 6.5 /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 80 v3.25.0 97 v3.25.0 97 v3.25.0 97 v3.25.0 基准（参考）

雷达图例

共 20 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 档位思考量 tier_thinking_volume_match
20 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 8405 ms 9437 ms 15353 ms 18286 ms 10275 ms

输出速度 178.4 tok/s 171.0 tok/s 134.9 tok/s 130.5 tok/s 143.1 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 ⓘ

$0.0884

$0.0986

$0.1115

$0.1084

$0.1336

缓存读取占比 ⓘ 70% 70% 70% 70% 47%

揭晓

通道

服务商 AIMZ ↗

通道M-Max

服务商 LingxiCode ↗

通道M-

服务商 OmniaKey ↗

通道O-Max

服务商 duckcoding ↗

通道cc专用

官方基准

返回列表