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

0-0

O-Max

luminai

M-Kiro/Antigravity

LinkAPI

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 18:38:41 ~ 18:38:56 (12.6s) 18:38:43 ~ 18:38:58 (11.8s) 18:38:45 ~ 18:38:50 (4.2s) 18:38:47 ~ 18:38:52 (4.8s) 18:38:49 ~ 18:38:54 (4.2s)

提示词

查看本步 Prompt

ping. Reply with the single word: pong

HTTP 头(TTFB) 8,597 ms 9,317 ms 2,087 ms 2,203 ms 2,040 ms

首文本(SSE delta) 8,599 ms 9,348 ms 2,107 ms 2,205 ms 2,041 ms

总耗时 10,967 ms 9,401 ms 2,581 ms 3,026 ms 2,598 ms

输入 10 1,443 10 10 10

输出 61 1 60 56 50

缓存创建

33,192

5m: 33,192 / 1h: 0

4,111

5m: 4,111 / 1h: 0

32,406

5m: 32,406 / 1h: 0

33,122

5m: 0 / 1h: 33,122

33,122

5m: 0 / 1h: 33,122

缓存读取 0 23,301 0 0 0

输出速度 25.8 tok/s - 126.6 tok/s 68.2 tok/s 89.8 tok/s

响应

pong

pong

pong

pong

pong

步骤 2: identity text

时间 18:42:38 ~ 18:42:53 (10.0s) 18:42:29 ~ 18:42:39 (5.3s) 18:40:45 ~ 18:40:51 (4.7s) 18:39:59 ~ 18:40:09 (7.1s) 18:42:36 ~ 18:42:46 (5.2s)

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) 7,360 ms 3,387 ms 2,140 ms 2,297 ms 1,718 ms

首文本(SSE delta) 7,361 ms 3,388 ms 2,162 ms 2,303 ms 1,720 ms

总耗时 8,455 ms 3,717 ms 3,667 ms 6,003 ms 4,194 ms

输入 3 1,452 10 10 10

输出 26 28 176 266 210

缓存创建

7,629

5m: 7,629 / 1h: 0

4,137

5m: 4,137 / 1h: 0

189

5m: 189 / 1h: 0

247

5m: 0 / 1h: 247

241

5m: 0 / 1h: 241

缓存读取 25,629 23,446 32,406 33,122 33,122

输出速度 - - 116.9 tok/s 71.9 tok/s 84.9 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 18:45:51 ~ 18:45:56 (4.5s) 18:43:40 ~ 18:43:50 (5.9s) 18:44:14 ~ 18:44:19 (4.1s) 18:41:29 ~ 18:41:34 (4.2s) 18:44:09 ~ 18:44:14 (3.4s)

提示词

查看本步 Prompt

Knowledge cutoff date? Reply with YYYY-MM only.

HTTP 头(TTFB) 2,682 ms 4,647 ms 2,367 ms 2,101 ms 1,698 ms

首文本(SSE delta) 2,710 ms 4,680 ms 2,392 ms 2,102 ms 1,701 ms

总耗时 2,758 ms 4,737 ms 2,842 ms 3,057 ms 2,335 ms

输入 3 1,455 10 10 10

输出 8 4 62 71 73

缓存创建

5m: 42 / 1h: 0

4,146

5m: 4,146 / 1h: 0

5m: 59 / 1h: 0

289

5m: 0 / 1h: 289

233

5m: 0 / 1h: 233

缓存读取 33,258 23,498 32,595 33,369 33,363

输出速度 - - 137.8 tok/s 74.3 tok/s 115.1 tok/s

响应

2025-02

2025-02

2025-02

2025-02

2025-02

步骤 4: identity_free text

时间 18:48:58 ~ 18:49:08 (7.9s) 18:47:07 ~ 18:47:17 (5.8s) 18:47:23 ~ 18:47:33 (6.8s) 18:44:37 ~ 18:44:47 (5.3s) 18:48:06 ~ 18:48:16 (5.3s)

提示词

查看本步 Prompt

用中文说说你的身份：什么产品？由谁打造？50 字以内。

HTTP 头(TTFB) 4,795 ms 3,887 ms 2,275 ms 1,597 ms 2,008 ms

首文本(SSE delta) 5,133 ms 4,435 ms 2,291 ms 1,598 ms 2,010 ms

总耗时 6,372 ms 4,435 ms 5,749 ms 4,282 ms 4,019 ms

输入 3 1,457 10 10 10

输出 66 52 554 229 375

缓存创建

5m: 39 / 1h: 0

4,151

5m: 4,151 / 1h: 0

5m: 35 / 1h: 0

113

5m: 0 / 1h: 113

107

5m: 0 / 1h: 107

缓存读取 33,300 23,527 32,654 33,658 33,596

输出速度 53.3 tok/s burst 160.2 tok/s 85.3 tok/s 186.7 tok/s

响应

我是由 Anthropic 公司打造的 Claude Haiku 4.5，一个大型语言模型。我可以帮助用户进行代码编写、文本分析、问题解答等多种任务。

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编程、分析、写作等各种任务。

我是Claude，由Anthropic公司开发的AI助手。当前版本是Claude Haiku 4.5，可以进行代码编辑、分析和问答。

我是Claude Haiku 4.5，由美国AI公司Anthropic开发的对话式AI助手。我可以帮助用户进行编程、分析、写作、数学计算等多种任务。

我是Claude，由Anthropic公司开发的AI助手。我可以帮助用户进行对话、编写和分析代码、回答问题等多种任务。

步骤 5: knowledge_recall text

时间 18:52:26 ~ 18:52:36 (8.3s) 18:48:53 ~ 18:49:03 (8.9s) 18:49:35 ~ 18:49:55 (17.9s) 18:48:41 ~ 18:49:06 (23.0s) 18:50:20 ~ 18:50:40 (19.6s)

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) 3,974 ms 6,329 ms 1,729 ms 1,736 ms 1,555 ms

首文本(SSE delta) 3,976 ms 6,330 ms 1,749 ms 1,737 ms 1,557 ms

总耗时 6,728 ms 7,538 ms 16,776 ms 22,062 ms 18,639 ms

输入 3 375 10 10 10

输出 203 192 2,535 2,081 2,639

缓存创建

8,094

5m: 0 / 1h: 8,094

1,067

5m: 1,067 / 1h: 0

8,067

5m: 8,067 / 1h: 0

8,618

5m: 0 / 1h: 8,618

8,696

5m: 0 / 1h: 8,696

缓存读取 0 6,048 0 0 0

输出速度 73.8 tok/s 158.9 tok/s 168.7 tok/s 102.4 tok/s 154.5 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 18:55:36 ~ 18:56:11 (34.4s) 18:51:16 ~ 18:51:21 (3.3s) 18:52:32 ~ 18:53:02 (29.5s) 18:51:19 ~ 18:51:39 (17.4s) 18:53:36 ~ 18:54:11 (33.1s)

提示词

查看本步 Prompt

Count digits exactly in s. Return {"n":len(s),"c":[count0,count7,count9]}. s=604304657568444944143039341023835123776524074411369370107603261550190780

HTTP 头(TTFB) 3,375 ms 2,014 ms 980 ms 1,394 ms 890 ms

首文本(SSE delta) 3,376 ms 2,015 ms 990 ms 1,398 ms 896 ms

总耗时 33,196 ms 2,250 ms 28,368 ms 16,458 ms 32,073 ms

输入 193 188 193 193 193

输出 5,749 19 2,601 2,791 6,421

缓存创建

5m: 0 / 1h: 0

缓存读取 0 - 0 0 0

输出速度 192.8 tok/s - 95.0 tok/s 185.3 tok/s 206.0 tok/s

响应

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

```json
{"n":72,"c":[14,9,8]}
```

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

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

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

总计

总输入(含缓存) 141,398 123,802 138,654 142,781 142,723

总输出 6,113 296 5,988 5,494 9,768

总缓存创建

48,996

5m: 40,902 / 1h: 8,094

17,612

5m: 17,612 / 1h: 0

40,756

5m: 40,756 / 1h: 0

42,389

5m: 0 / 1h: 42,389

42,399

5m: 0 / 1h: 42,399

总缓存读取 92,187 99,820 97,655 100,149 100,081

输出速度 168.0 tok/s 158.9 tok/s 124.0 tok/s 126.2 tok/s 181.1 tok/s

总执行时间 1m 17.6s 41.0s 1m 7.0s 1m 1.8s 1m 10.7s

总墙钟时间

17m 30.0s

含步骤间等待

12m 38.0s

含步骤间等待

14m 17.1s

含步骤间等待

12m 52.0s

含步骤间等待

15m 22.1s

含步骤间等待

按官方价目重估

同等内容若直发 Anthropic 的估算 ⓘ

$0.0821

Anthropic: Claude Haiku 4.5

$0.0316

Anthropic: Claude Haiku 4.5

$0.0718

Anthropic: Claude Haiku 4.5

$0.0907

Anthropic: Claude Haiku 4.5

$0.1121

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_014CpnN48FbxmhP52XXv9ids msg_018ZhSbsUqRr3i5rHtkrTCBAFr msg_016RWVVSuVs5wFWnFWDM5GuA msg_011z5RGic2rQDK9QX53R7r8h msg_01EhTqUcSZryaXBZ1YA3i3Tt

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	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	a3ff438495c16569df14689e906d6763
M-Kiro/Antigravity	cf-ray: a0ea37ac4a7166d3-AMS server: cloudflare x-oneapi-request-id: 202606201051182918317398268d9d6uyRtTvoA	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	a0ea37ac4a7166d3-AMS
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: 202606201052338535443698268d9d6lxdPAbvq	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	—
cc专用	server: Nginx x-oneapi-request-id: 202606201051207103173858268d9d699fIgJTC	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_011CcEPQjognKtJdmhWGZkhU

协议指纹评分（机器自动）

总分 / 100 85 69 88 97 100

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

缓存 TTL 一致 0.0 /10 ×15 0.0 /10 ×15 0.0 /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 5.0 /10 ×8 10.0 /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 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 0.0 /10 ×5 10.0 /10 ×5 10.0 /10 ×5 10.0 /10 ×5

延迟基线 6.1 /10 ×5 7.8 /10 ×5 10.0 /10 ×5 10.0 /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 85 v3.24.0 69 v3.24.0 88 v3.24.0 97 v3.24.0 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 信封自报一致 envelope_self_report_match
20 中文遵循 instruction_following_lang
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 11413 ms 5346 ms 9997 ms 9148 ms 10643 ms

输出速度 168.0 tok/s 158.9 tok/s 124.0 tok/s 126.2 tok/s 181.1 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 ⓘ

$0.0821

$0.0316

$0.0718

$0.0907

$0.1121

缓存读取占比 ⓘ 65% 81% 70% 70% 70%

揭晓

通道

服务商 0-0 ↗

通道O-Max

服务商 luminai ↗

通道M-Kiro/Antigravity

服务商 LinkAPI ↗

通道O-Max

服务商 duckcoding ↗

通道cc专用

官方基准

返回列表