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

1 任务失败执行不完整

已揭晓揭晓进度

步骤 / 指标

SAIAi

O-Max

InfAI

R-Kiro

官方基准

SSSAiCode

O-Max-L2

AiTokensFlux

O-Pro

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 09:30:59 ~ 09:31:04 (3.1s) 09:30:56 ~ 09:31:07 (5.1s) 09:31:01 ~ 09:31:06 (2.9s) 09:30:53 ~ 09:30:58 (3.8s) 09:30:54 ~ 09:31:05 (8.6s)

提示词

查看本步 Prompt

ping. Reply with the single word: pong

HTTP 头(TTFB) 1,644 ms 3,535 ms 1,568 ms 2,490 ms 4,694 ms

首文本(SSE delta) 1,646 ms 3,537 ms 1,570 ms 2,899 ms 4,717 ms

总耗时 2,056 ms 3,537 ms 1,882 ms 3,041 ms 5,505 ms

输入 10 10 10 10 10

输出 43 41 38 47 55

缓存创建

33,122

5m: 0 / 1h: 33,122

33,122

5m: 0 / 1h: 33,122

5m: 0 / 1h: 0

32,787

5m: 0 / 1h: 32,787

33,261

5m: 0 / 1h: 33,261

缓存读取 0 0 33,122 0 0

输出速度 104.9 tok/s burst 121.8 tok/s - 69.8 tok/s

响应

pong

pong

pong

pong

pong

步骤 2: identity text

时间 09:35:01 ~ 09:35:11 (5.1s) 09:32:23 ~ 09:32:33 (6.3s) 09:32:18 ~ 09:32:23 (4.8s) 09:32:35 ~ 09:32:45 (7.6s) 09:34:39 ~ 09:34:54 (13.1s)

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) 1,862 ms 2,993 ms 1,544 ms 2,275 ms 10,355 ms

首文本(SSE delta) 1,863 ms 2,994 ms 1,545 ms 6,451 ms 10,356 ms

总耗时 4,094 ms 4,895 ms 3,741 ms 6,742 ms 12,321 ms

输入 10 10 10 10 10

输出 227 299 208 425 194

缓存创建

237

5m: 0 / 1h: 237

235

5m: 0 / 1h: 235

232

5m: 0 / 1h: 232

241

5m: 0 / 1h: 241

1,884

5m: 0 / 1h: 1,884

缓存读取 33,122 33,122 33,122 32,787 31,626

输出速度 101.7 tok/s 157.3 tok/s 94.7 tok/s burst 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-3-5-sonnet-20241022

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

步骤 3: cutoff text

时间 09:39:11 ~ 09:39:16 (3.5s) 09:35:45 ~ 09:35:55 (5.0s) 09:35:03 ~ 09:35:08 (4.5s) 09:34:25 ~ 09:34:30 (2.4s) 09:35:54 ~ 09:35:59 (4.0s)

提示词

查看本步 Prompt

Knowledge cutoff date? Reply with YYYY-MM only.

HTTP 头(TTFB) 1,860 ms 3,704 ms 2,433 ms 1,053 ms 2,399 ms

首文本(SSE delta) 1,862 ms 3,705 ms 2,434 ms 1,540 ms 2,401 ms

总耗时 2,457 ms 3,705 ms 3,493 ms 1,699 ms 3,243 ms

输入 10 10 10 10 10

输出 63 74 82 84 78

缓存创建

250

5m: 0 / 1h: 250

322

5m: 0 / 1h: 322

250

5m: 0 / 1h: 250

467

5m: 0 / 1h: 467

485

5m: 0 / 1h: 485

缓存读取 33,359 33,357 33,354 33,028 33,261

输出速度 105.9 tok/s burst 77.4 tok/s - 92.6 tok/s

响应

2025-02

2025-02

2025-02

2024-10

2025-02

步骤 4: identity_free text

时间 09:41:18 ~ 09:41:28 (5.8s) 09:39:34 ~ 09:39:49 (10.0s) 09:38:01 ~ 09:38:16 (14.3s) 09:36:31 ~ 09:36:36 (3.0s) 09:38:43 ~ 09:38:53 (5.9s)

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) 1,284 ms 4,133 ms 1,648 ms 1,120 ms 3,282 ms

首文本(SSE delta) 1,285 ms 4,134 ms 1,650 ms 1,908 ms 3,287 ms

总耗时 4,790 ms 8,506 ms 13,322 ms 2,051 ms 5,027 ms

输入 10 10 10 10 10

输出 418 823 1,307 152 154

缓存创建

101

5m: 0 / 1h: 101

112

5m: 0 / 1h: 112

120

5m: 0 / 1h: 120

124

5m: 0 / 1h: 124

112

5m: 0 / 1h: 112

缓存读取 33,609 33,679 33,604 33,495 33,746

输出速度 119.3 tok/s 188.2 tok/s 112.0 tok/s burst 88.5 tok/s

响应

我是Claude Haiku 4.5，由Anthropic公司打造的大语言模型。我是Claude家族中的轻量级版本，具备快速响应和高效推理的特点。

我是Claude Haiku 4.5，由Anthropic公司开发的AI助手。Haiku是Claude家族中最轻量高效的模型，专为快速交互任务优化设计。

我是Claude Haiku，由Anthropic公司打造的AI助手。Haiku是Claude系列中最轻量的模型，面向高速和低成本场景优化。

我叫Claude，是Anthropic公司开发的AI助手。属于Claude产品系列，当前版本为Claude 3.5 Sonnet。

我是Claude，由Anthropic公司开发的AI助手。我可以帮助用户进行编程、写作、分析等各种任务，使用最新的Claude模型提供高质量的回答。

步骤 5: knowledge_recall text

时间 09:44:16 ~ 09:44:36 (15.8s) 09:41:06 ~ 09:41:51 (41.7s) 09:40:29 ~ 09:40:54 (22.9s) 09:40:34 ~ 09:40:49 (12.6s) 09:42:41 ~ 09:43:01 (15.5s)

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) 1,451 ms 2,468 ms 1,454 ms 1,104 ms 1,960 ms

首文本(SSE delta) 1,453 ms 2,469 ms 1,455 ms 11,401 ms 1,961 ms

总耗时 14,828 ms 40,540 ms 21,868 ms 11,724 ms 14,715 ms

输入 10 10 10 10 10

输出 1,676 6,054 2,417 1,758 2,249

缓存创建

3,025

5m: 0 / 1h: 3,025

9,232

5m: 0 / 1h: 9,232

9,649

5m: 0 / 1h: 9,649

8,389

5m: 0 / 1h: 8,389

8,491

5m: 0 / 1h: 8,491

缓存读取 5,721 0 0 0 0

输出速度 125.3 tok/s 159.0 tok/s 118.4 tok/s burst 176.3 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

{"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

```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

步骤 6: digit_count text

时间 09:47:38 ~ 09:48:09 (29.5s) 09:45:34 ~ 09:45:49 (13.1s) 09:43:38 ~ 09:44:13 (33.1s) 09:44:37 ~ 09:44:53 (14.2s) 09:44:33 ~ 09:44:38 (1.0s)

提示词

查看本步 Prompt

Count digits exactly in s. Return {"n":len(s),"c":[count0,count7,count9]}. s=503209952156584854423931425556960514655573311931051880236823042664353094

HTTP 头(TTFB) 874 ms 1,669 ms 1,422 ms 917 ms -

首文本(SSE delta) 876 ms 1,670 ms 1,423 ms 924 ms -

总耗时 28,475 ms 11,985 ms 32,195 ms 13,421 ms -

输入 193 193 193 139 -

输出 3,389 2,029 4,076 2,202 -

缓存创建

5m: 0 / 1h: 0

缓存读取 0 0 0 0 -

输出速度 122.8 tok/s 196.7 tok/s 132.5 tok/s 176.2 tok/s -

响应

```json
{"n":72,"c":[7,1,6]}
```

```json
{"n":72,"c":[7,1,6]}
```

```json
{"n": 71, "c": [7, 1, 6]}
```

```json
{"n":72,"c":[7,1,6]}
```

失败

评测失败

总计

总输入(含缓存)

142,789

143,424

143,696

141,507

142,916

仅前 6 步

总输出

5,816

9,320

8,128

4,668

2,730

仅前 6 步

总缓存创建

36,735

5m: 0 / 1h: 36,735

43,023

5m: 0 / 1h: 43,023

10,251

5m: 0 / 1h: 10,251

42,008

5m: 0 / 1h: 42,008

44,233

5m: 0 / 1h: 44,233

仅前 6 步

总缓存读取

105,811

100,158

133,202

99,310

98,633

仅前 6 步

输出速度

121.9 tok/s

168.4 tok/s

122.4 tok/s

176.2 tok/s

150.9 tok/s

仅前 6 步

总执行时间

1m 2.8s

1m 21.3s

1m 22.6s

43.5s

48.1s

仅前 6 步

总墙钟时间

17m 10.0s

含步骤间等待

14m 52.1s

含步骤间等待

13m 12.8s

含步骤间等待

14m 0.0s

含步骤间等待

13m 44.0s

含步骤间等待

仅前 6 步

按官方价目重估

同等内容若直发 Anthropic 的估算 ⓘ

$0.0858

Anthropic: Claude Haiku 4.5

$0.1106

Anthropic: Claude Haiku 4.5

$0.0670

Anthropic: Claude Haiku 4.5

$0.0860

Anthropic: Claude Haiku 4.5

$0.0789

Anthropic: Claude Haiku 4.5 (部分)

通道指纹

来源步骤 digit_count digit_count digit_count digit_count knowledge_recall

平台 direct (100%) direct (100%) direct (100%) direct (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 - - - - -

ID 格式 msg_01V4EVNpiy9gcCSbTo4juBdj msg_01B2KJcE1qREs2YzgZwsfoxX msg_01VrLq1LRm1Cg1xbMnXhji7c msg_01QHTLk6QMzu6DXMbQJNWU1G msg_01PnWgYBYAQaq1biZBZ5qjcM

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_011CcWi4FXJTLkAi4gEEXzWz
R-Kiro	server: nginx	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	—
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_011CcWhkabWsPC2ZueBkAEC4
O-Max-L2	cf-ray: a1313f4b0b93c8d2-LAX server: Photon-Edge set-cookie: _cfuvid=sF3W3TNxk7kE8.ELmXZB75FfHd9KrLea2GkdxXiYok0-1782697478.8913682-1.0.1.1-_LHfbN3xk1x4xt0kt.ivtDFxA7VIITyZenSMK37frow; HttpOnly; SameSite=None; Secure; Path=/; Domain=api.anthropic.com	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	1782697478787-ofxm1top36ar1 req_011CcWhpuBpLqBL6T5bpChrx a1313f4b0b93c8d2-LAX
O-Pro	server: nginx	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	—

协议指纹评分（机器自动）

总分 / 100 100 97 99 90

基线状态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 —

缓存 TTL 一致 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 —

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 —

原生 msg-ID 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 10.0 /10 ×8 10.0 /10 ×8 5.0 /10 ×8 —

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 5.0 /10 ×7 —

身份(自由) 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 —

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

延迟基线 10.0 /10 ×5 8.0 /10 ×5 10.0 /10 ×5 9.8 /10 ×5 —

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

stop_reason 10.0 /10 ×3 10.0 /10 ×3 10.0 /10 ×3 10.0 /10 ×3 —

SDK 一致 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 —

综合结论

与基线相似度 100 v3.25.0 97 v3.25.0 基准（参考） 90 v3.25.0 执行失败，不评分

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 9450 ms 12195 ms 12750 ms 6446 ms 8162 ms

输出速度 121.9 tok/s 168.4 tok/s 122.4 tok/s 176.2 tok/s 150.9 tok/s

完成步骤 6/6 成功 6/6 成功 6/6 成功 6/6 成功 5/6 成功 · 1 失败

按官方价目重估

同等内容若直发 Anthropic 的估算 ⓘ

$0.0858

$0.1106

$0.0670

$0.0860

$0.0789 (部分)

缓存读取占比 ⓘ 74% 70% 93% 70% 69% (部分)

揭晓

通道

服务商 SAIAi ↗

通道O-Max

服务商 InfAI ↗

通道R-Kiro

官方基准

服务商 SSSAiCode ↗

通道O-Max-L2

服务商 AiTokensFlux ↗

通道O-Pro

返回列表