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

2026/06/20 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

SAIAi

O-Max

HongMaCC

u-cc

SudoCode

M-Max

AIMZ

M-Max

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 13:54:00 ~ 13:54:05 (4.3s) 13:53:56 ~ 13:54:01 (4.8s) 13:53:58 ~ 13:54:08 (5.2s) 13:53:54 ~ 13:54:04 (8.4s) 13:54:02 ~ 13:54:07 (3.7s)

提示词

查看本步 Prompt

ping. Reply with the single word: pong

HTTP 头(TTFB) 2,588 ms 3,373 ms 3,199 ms 2,346 ms 1,986 ms

首文本(SSE delta) 2,591 ms 3,375 ms 3,201 ms 2,611 ms 1,988 ms

总耗时 3,124 ms 3,376 ms 3,665 ms 3,197 ms 2,661 ms

输入 10 10 10 10 10

输出 61 54 53 45 52

缓存创建

33,122

5m: 0 / 1h: 33,122

29,188

5m: 0 / 1h: 29,188

33,122

5m: 0 / 1h: 33,122

32,785

5m: 0 / 1h: 32,785

33,122

5m: 0 / 1h: 33,122

缓存读取 0 0 0 0 0

输出速度 114.4 tok/s burst 114.2 tok/s 76.8 tok/s 77.3 tok/s

响应

pong

pong

pong

pong

pong

步骤 2: identity text

时间 13:57:56 ~ 13:58:06 (6.8s) 13:55:02 ~ 13:55:12 (6.8s) 13:55:10 ~ 13:55:20 (7.1s) 13:56:32 ~ 13:56:42 (8.3s) 13:55:14 ~ 13:55:24 (6.5s)

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) 4,133 ms 5,682 ms 3,687 ms 2,086 ms 1,775 ms

首文本(SSE delta) 4,135 ms 5,683 ms 3,689 ms 6,178 ms 1,777 ms

总耗时 5,870 ms 5,684 ms 6,015 ms 6,391 ms 5,306 ms

输入 10 10 10 10 10

输出 237 223 333 471 267

缓存创建

265

5m: 0 / 1h: 265

202

5m: 0 / 1h: 202

257

5m: 0 / 1h: 257

249

5m: 0 / 1h: 249

256

5m: 0 / 1h: 256

缓存读取 33,122 29,188 33,122 32,785 33,122

输出速度 136.6 tok/s burst 143.2 tok/s burst 75.7 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-3-5-sonnet

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

步骤 3: cutoff text

时间 14:01:59 ~ 14:02:04 (3.4s) 13:58:46 ~ 13:58:51 (4.7s) 13:57:39 ~ 13:57:44 (4.3s) 13:58:55 ~ 13:59:00 (3.3s) 13:57:31 ~ 13:57:36 (3.7s)

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) 1,987 ms 3,509 ms 2,600 ms 1,878 ms 1,650 ms

首文本(SSE delta) 1,989 ms 3,510 ms 2,602 ms 2,258 ms 1,654 ms

总耗时 2,438 ms 3,510 ms 3,239 ms 2,306 ms 2,586 ms

输入 10 10 10 10 10

输出 76 82 93 58 74

缓存创建

279

5m: 0 / 1h: 279

5m: 0 / 1h: 61

375

5m: 0 / 1h: 375

513

5m: 0 / 1h: 513

309

5m: 0 / 1h: 309

缓存读取 33,387 29,390 33,379 33,034 33,378

输出速度 169.3 tok/s burst 146.0 tok/s burst 79.4 tok/s

响应

2025-02

2025-02

2025-02

2024-04

2025-02

步骤 4: identity_free text

时间 14:03:19 ~ 14:03:29 (5.1s) 14:00:47 ~ 14:01:07 (17.2s) 13:59:00 ~ 13:59:10 (9.9s) 14:01:34 ~ 14:01:49 (12.1s) 14:00:39 ~ 14:00:54 (11.6s)

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) 1,718 ms 15,991 ms 2,602 ms 3,726 ms 1,868 ms

首文本(SSE delta) 1,722 ms 15,993 ms 2,604 ms 9,380 ms 1,870 ms

总耗时 3,975 ms 15,993 ms 8,895 ms 9,878 ms 10,617 ms

输入 10 10 10 10 10

输出 466 1,282 326 680 783

缓存创建

118

5m: 0 / 1h: 118

5m: 0 / 1h: 39

127

5m: 0 / 1h: 127

100

5m: 0 / 1h: 100

116

5m: 0 / 1h: 116

缓存读取 33,666 29,451 33,754 33,547 33,687

输出速度 206.8 tok/s burst 51.8 tok/s burst 89.5 tok/s

响应

我是Claude，由Anthropic开发的AI助手。是一个大语言模型，可以帮助用户编程、写作、分析和解决各种问题。

我是Anthropic开发的Claude AI助手，当前模型为Claude Haiku 4.5，用于对话、编程和内容创作。

我是Claude Haiku 4.5，由Anthropic公司开发的AI助手。我可以帮助进行编程、代码审查、软件设计等各种技术任务。

我是Claude，由Anthropic开发的AI助手。可帮助编程、分析、写作、研究等多种任务。

我是Claude，由Anthropic开发的AI编程助手。基于Haiku 4.5模型，可帮助代码开发、调试、重构等软件工程任务。

步骤 5: knowledge_recall text

时间 14:06:22 ~ 14:06:32 (9.8s) 14:03:22 ~ 14:03:52 (26.9s) 14:03:08 ~ 14:03:33 (21.4s) 14:04:12 ~ 14:04:27 (11.1s) 14:04:53 ~ 14:05:43 (49.8s)

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) 1,619 ms 25,508 ms 2,249 ms 1,850 ms 1,882 ms

首文本(SSE delta) 1,622 ms 25,510 ms 2,250 ms 8,813 ms 1,884 ms

总耗时 8,855 ms 25,510 ms 20,479 ms 9,893 ms 48,846 ms

输入 10 10 10 10 10

输出 1,329 1,967 2,699 1,446 3,961

缓存创建

8,868

5m: 0 / 1h: 8,868

4,389

5m: 0 / 1h: 4,389

8,825

5m: 0 / 1h: 8,825

8,945

5m: 0 / 1h: 8,945

9,204

5m: 0 / 1h: 9,204

缓存读取 0 11,959 0 0 0

输出速度 183.7 tok/s burst 148.1 tok/s burst 84.3 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

{"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

```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 14:09:43 ~ 14:10:08 (20.7s) 14:06:00 ~ 14:06:25 (20.4s) 14:06:53 ~ 14:07:08 (14.5s) 14:08:13 ~ 14:08:38 (23.7s) 14:08:37 ~ 14:09:02 (21.0s)

提示词

查看本步 Prompt

Count digits exactly in s. Return {"n":len(s),"c":[count0,count7,count9]}. s=714873388778255288384361821339017462051493419852738413751897729945583226

HTTP 头(TTFB) 851 ms 19,243 ms 1,436 ms 2,324 ms 981 ms

首文本(SSE delta) 852 ms 19,244 ms 1,454 ms 2,369 ms 983 ms

总耗时 19,570 ms 19,244 ms 13,386 ms 22,400 ms 20,037 ms

输入 193 10 193 139 193

输出 3,444 2,427 2,090 3,603 3,674

缓存创建

5m: 0 / 1h: 0

2,193

5m: 0 / 1h: 2,193

5m: 0 / 1h: 0

缓存读取 0 11,959 0 0 0

输出速度 184.0 tok/s burst 175.2 tok/s 179.9 tok/s 192.8 tok/s

响应

```json
{"n": 72, "c": [2, 9, 6]}
```

I'll count the digits in the string s.

```
s = "714873388778255288384361821339017462051493419852738413751897729945583226"
```

Let me count:
- **Total length (n):** 72 characters
- **Count of '0':**

```json
{"n": 72, "c": [2, 9, 6]}
```

```json
{"n":72,"c":[2,9,6]}
```

```json
{"n":72,"c":[2,9,6]}
```

总计

总输入(含缓存) 143,070 148,079 143,204 142,147 143,437

总输出 5,613 6,035 5,594 6,303 8,811

总缓存创建

42,652

5m: 0 / 1h: 42,652

36,072

5m: 0 / 1h: 36,072

42,706

5m: 0 / 1h: 42,706

42,592

5m: 0 / 1h: 42,592

43,007

5m: 0 / 1h: 43,007

总缓存读取 100,175 111,947 100,255 99,366 100,187

输出速度 181.5 tok/s burst 140.3 tok/s 176.9 tok/s 110.3 tok/s

总执行时间 50.2s 1m 20.9s 1m 2.5s 1m 6.9s 1m 36.3s

总墙钟时间

16m 8.0s

含步骤间等待

12m 29.0s

含步骤间等待

13m 10.0s

含步骤间等待

14m 44.0s

含步骤间等待

15m 0.1s

含步骤间等待

按官方价目重估

同等内容若直发 Anthropic 的估算 ⓘ

$0.0916

Anthropic: Claude Haiku 4.5

$0.0865

Anthropic: Claude Haiku 4.5

$0.0916

Anthropic: Claude Haiku 4.5

$0.0949

Anthropic: Claude Haiku 4.5

$0.1081

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 direct (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 -

ID 格式 msg_0175sBb2kFHT5Nz53PQsVSc2 msg_01TgSfwVLcA5NkziPbdAYREP msg_019b7xxnciJdNRetJWUBgaTx msg_012jqWKJwsafC32jsqyk76o2 msg_01GJJsXUe6gfPrN8cgnDLCDg

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	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_011CcE1m8KGosh5SPWkTP7Rq
u-cc	server: nginx x-new-api-version: 1.0.14 x-oneapi-request-id: 202606200606014624924678268d9d6kP627HKC	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-idbody:id:msg_prefixbody:usage.service_tierbody:usage.inference_geobody:model:date_suffixbody:usage.cache_fieldsbody:extra_fieldsupstream:anthropic-direct-from-body	—
M-Max	server: nginx x-new-api-version: v1.0.0-rc.11 x-oneapi-request-id: 202606201406544443383081kyznzkp	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-idbody:id:msg_prefixbody:usage.service_tierbody:usage.inference_geobody:model:date_suffixbody:usage.cache_fieldsbody:extra_fieldsupstream:anthropic-direct-from-body	—
M-Max	cf-ray: a0e8990b2b3b9deb-AMS server: cloudflare x-new-api-version: v0.0.0 x-oneapi-request-id: 202606200608146268006338268d9d6qz8eVm7w	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	a0e8990b2b3b9deb-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_011CcE1gGS2xkWzmWUFeDq1x

协议指纹评分（机器自动）

总分 / 100 98 90 98 81 99

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 10.0 /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 10.0 /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 10.0 /10 ×8 0.0 /10 ×8 10.0 /10 ×8 5.0 /10 ×8 10.0 /10 ×8

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

知识截止 10.0 /10 ×7 10.0 /10 ×7 10.0 /10 ×7 0.0 /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 10.0 /10 ×5 9.8 /10 ×5 10.0 /10 ×5 10.0 /10 ×5 10.0 /10 ×5

Req-ID 透传 10.0 /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 v3.24.0 90 v3.24.0 98 v3.24.0 81 v3.24.0 基准（参考）

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 7305 ms 12220 ms 9280 ms 9011 ms 15009 ms

输出速度 181.5 tok/s burst 140.3 tok/s 176.9 tok/s 110.3 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 ⓘ

$0.0916

$0.0865

$0.0916

$0.0949

$0.1081

缓存读取占比 ⓘ 70% 76% 70% 70% 70%

揭晓

通道

服务商 SAIAi ↗

通道O-Max

服务商 HongMaCC ↗

通道u-cc

服务商 SudoCode ↗

通道M-Max

服务商 AIMZ ↗

通道M-Max

官方基准

返回列表