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

全部 ≥ 70 与基线相似度

已揭晓揭晓进度

步骤 / 指标

InfAI

R-Kiro

ByteCatCode

O-Max

官方基准

Novaris

O-Max

ClaudeZZ

R-kiro

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 08:14:00 ~ 08:14:05 (4.2s) 08:13:55 ~ 08:14:00 (3.9s) 08:14:02 ~ 08:14:07 (3.5s) 08:13:57 ~ 08:14:12 (11.7s) 08:13:53 ~ 08:14:03 (5.1s)

提示词

查看本步 Prompt

ping. Reply with the single word: pong

HTTP 头(TTFB) 2,900 ms 2,075 ms 2,008 ms 2,964 ms 3,229 ms

首文本(SSE delta) 2,901 ms - 2,014 ms 2,970 ms 3,230 ms

总耗时 2,901 ms 2,632 ms 2,401 ms 3,432 ms 3,971 ms

输入 10 10 10 10 10

输出 37 49 47 62 52

缓存创建

33,122

5m: 0 / 1h: 33,122

33,122

5m: 0 / 1h: 33,122

33,122

5m: 0 / 1h: 33,122

33,122

5m: 0 / 1h: 33,122

22,405

5m: 0 / 1h: 22,405

缓存读取 0 0 0 0 0

输出速度 burst 88.0 tok/s 121.4 tok/s 134.2 tok/s 70.2 tok/s

响应

pong

pong

pong

pong

pong

步骤 2: identity text

时间 08:17:02 ~ 08:17:12 (5.3s) 08:15:20 ~ 08:15:30 (5.8s) 08:15:31 ~ 08:15:46 (5.1s) 08:16:07 ~ 08:16:17 (5.8s) 08:17:17 ~ 08:17:27 (4.9s)

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) 3,179 ms 2,251 ms 1,510 ms 2,521 ms 2,633 ms

首文本(SSE delta) 3,180 ms - 1,512 ms 2,535 ms 2,635 ms

总耗时 4,018 ms 4,485 ms 4,049 ms 4,944 ms 3,978 ms

输入 10 10 10 10 10

输出 179 272 209 199 176

缓存创建

241

5m: 0 / 1h: 241

253

5m: 0 / 1h: 253

251

5m: 0 / 1h: 251

266

5m: 0 / 1h: 266

263

5m: 0 / 1h: 263

缓存读取 33,122 33,122 33,122 33,122 22,405

输出速度 213.6 tok/s 121.8 tok/s 82.4 tok/s 82.6 tok/s 131.0 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

步骤 3: cutoff text

时间 08:19:49 ~ 08:19:54 (3.9s) 08:19:20 ~ 08:19:25 (3.6s) 08:18:24 ~ 08:18:29 (4.5s) 08:18:21 ~ 08:18:26 (3.5s) 08:21:26 ~ 08:21:31 (4.5s)

提示词

查看本步 Prompt

你掌握的世界知识截止于何时？YYYY-MM 格式作答，不解释。

HTTP 头(TTFB) 2,752 ms 1,710 ms 2,310 ms 2,181 ms 3,013 ms

首文本(SSE delta) 2,753 ms - 2,312 ms 2,183 ms 3,014 ms

总耗时 2,753 ms 2,359 ms 3,435 ms 2,704 ms 3,725 ms

输入 10 10 10 10 10

输出 71 79 82 65 92

缓存创建

219

5m: 0 / 1h: 219

314

5m: 0 / 1h: 314

232

5m: 0 / 1h: 232

239

5m: 0 / 1h: 239

242

5m: 0 / 1h: 242

缓存读取 33,363 33,375 33,373 33,388 22,668

输出速度 burst 121.7 tok/s 73.0 tok/s 124.8 tok/s 129.4 tok/s

响应

2025-02

2025-02

2025-02

2025-02

2025-02

步骤 4: identity_free text

时间 08:23:29 ~ 08:23:54 (22.0s) 08:23:10 ~ 08:23:15 (4.3s) 08:21:56 ~ 08:22:16 (17.9s) 08:21:41 ~ 08:21:51 (7.9s) 08:25:12 ~ 08:25:27 (11.6s)

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) 4,897 ms 1,838 ms 1,590 ms 2,028 ms 3,372 ms

首文本(SSE delta) 4,898 ms - 1,591 ms 2,028 ms 3,373 ms

总耗时 20,681 ms 3,076 ms 16,895 ms 7,171 ms 10,813 ms

输入 10 10 10 10 10

输出 2,296 172 1,562 699 1,277

缓存创建

109

5m: 0 / 1h: 109

119

5m: 0 / 1h: 119

120

5m: 0 / 1h: 120

5m: 0 / 1h: 99

142

5m: 0 / 1h: 142

缓存读取 33,582 33,689 33,605 33,627 22,910

输出速度 145.5 tok/s 138.9 tok/s 102.1 tok/s 135.9 tok/s 171.6 tok/s

响应

我是Claude Haiku 4.5，由Anthropic开发。Claude系列中最轻量的模型，专为快速高效处理任务设计。

我叫Claude Haiku 4.5，属于Claude产品家族，由Anthropic公司开发。

我是Claude Haiku 4.5，由Anthropic公司开发的AI助手，支持对话、编码和分析，知识截止于2025年2月。

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 Haiku 4.5，由Anthropic开发，属于Claude产品系列，是高效的轻量级大语言模型。

步骤 5: knowledge_recall text

时间 08:24:57 ~ 08:25:12 (14.2s) 08:26:53 ~ 08:27:13 (18.9s) 08:24:01 ~ 08:25:16 (72.4s) 08:23:34 ~ 08:23:54 (19.7s) 08:29:00 ~ 08:29:30 (26.0s)

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) 3,028 ms 1,506 ms 2,663 ms 2,597 ms 2,980 ms

首文本(SSE delta) 3,029 ms - 2,695 ms 2,599 ms 2,982 ms

总耗时 12,968 ms 17,296 ms 71,331 ms 18,944 ms 25,132 ms

输入 10 10 10 10 10

输出 1,773 2,513 4,343 2,408 3,483

缓存创建

10,605

5m: 0 / 1h: 10,605

8,598

5m: 0 / 1h: 8,598

9,905

5m: 0 / 1h: 9,905

9,043

5m: 0 / 1h: 9,043

9,812

5m: 0 / 1h: 9,812

缓存读取 0 0 0 0 0

输出速度 178.4 tok/s 159.2 tok/s 63.3 tok/s 147.3 tok/s 157.2 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 08:28:33 ~ 08:28:53 (19.2s) 08:30:18 ~ 08:30:43 (21.8s) 08:26:38 ~ 08:27:38 (55.6s) 08:26:51 ~ 08:27:21 (29.8s) 08:30:44 ~ 08:31:04 (16.9s)

提示词

查看本步 Prompt

Count digits exactly in s. Return {"n":len(s),"c":[count0,count7,count9]}. s=770317297749407176739322825022146612969789684642684443710103484407507880

HTTP 头(TTFB) 3,227 ms 1,298 ms 993 ms 972 ms 1,604 ms

首文本(SSE delta) 3,228 ms - 994 ms 973 ms 1,622 ms

总耗时 16,061 ms 20,606 ms 54,518 ms 28,991 ms 16,121 ms

输入 193 193 193 193 10

输出 2,396 2,158 6,564 3,591 1,765

缓存创建

5m: 0 / 1h: 0

5,981

5m: 0 / 1h: 5,981

缓存读取 0 0 0 0 0

输出速度 186.7 tok/s 111.8 tok/s 122.6 tok/s 128.2 tok/s 121.7 tok/s

响应

```json
{"n":72,"c":[8,12,6]}
```

```json
{"n":72,"c":[8,12,6]}
```

```json
{"n": 72, "c": [8, 12, 6]}
```

```json
{"n":72,"c":[8,12,6]}
```

I'll count the digits in the string exactly.

```
s = "770317297749407176739322825022146612969789684642684443710103484407507880"
```

Counting:
- **Length (n)**: 72 characters
- **Count of '0'**: 8
-

总计

总输入(含缓存) 144,606 142,835 143,973 143,149 106,888

总输出 6,752 5,243 12,807 7,024 6,845

总缓存创建

44,296

5m: 0 / 1h: 44,296

42,406

5m: 0 / 1h: 42,406

43,630

5m: 0 / 1h: 43,630

42,769

5m: 0 / 1h: 42,769

38,845

5m: 0 / 1h: 38,845

总缓存读取 100,067 100,186 100,100 100,137 67,983

输出速度 168.7 tok/s 131.8 tok/s 90.5 tok/s 132.8 tok/s 146.0 tok/s

总执行时间 1m 8.8s 58.2s 2m 38.9s 1m 18.4s 1m 9.1s

总墙钟时间

14m 53.1s

含步骤间等待

16m 48.7s

含步骤间等待

13m 36.2s

含步骤间等待

13m 24.3s

含步骤间等待

17m 10.9s

含步骤间等待

按官方价目重估

同等内容若直发 Anthropic 的估算 ⓘ

$0.0994

Anthropic: Claude Haiku 4.5

$0.0895

Anthropic: Claude Haiku 4.5

$0.1288

Anthropic: Claude Haiku 4.5

$0.0988

Anthropic: Claude Haiku 4.5

$0.0896

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%) direct (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 - - - - -

ID 格式 msg_01Fpe8HAjZtVMv7SUYNPFNN8 msg_016BQj2DphcDKxey4K2U2ee7 msg_01M1oU4u6uTFhFGAXEAqZNvk msg_01KWSzQyFvzMK9AE1m4nWKCX msg_012uJhbh3hqpuuhxXUQPLtmC

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
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	—
O-Max	server: nginx/1.29.8 x-oneapi-request-id: 20260629003020647662228268d9d66gjk8jxN	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_011CcWbsynXKHqNckxiDqEch
O-Max	server: nginx x-oneapi-request-id: 202606290026521958614038268d9d6QwdnO8dr	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	—
R-kiro	via: 1.1 google cf-ray: c4539aa6dd7ed68f31890752789cda6a-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_c4539aa6dd7ed68f31890752789cda6a c4539aa6dd7ed68f31890752789cda6a-IAD

协议指纹评分（机器自动）

总分 / 100 97 97 99 96 92

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 10.0 /10 ×8 10.0 /10 ×8 10.0 /10 ×8 0.0 /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 10.0 /10 ×5 10.0 /10 ×5 10.0 /10 ×5 10.0 /10 ×5

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

Req-ID 透传 0.0 /10 ×4 0.0 /10 ×4 10.0 /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 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 97 v3.25.0 97 v3.25.0 基准（参考） 96 v3.25.0 92 v3.25.0

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 9897 ms 8409 ms 25438 ms 11031 ms 10623 ms

输出速度 168.7 tok/s 131.8 tok/s 90.5 tok/s 132.8 tok/s 146.0 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 ⓘ

$0.0994

$0.0895

$0.1288

$0.0988

$0.0896

缓存读取占比 ⓘ 69% 70% 70% 70% 64%

揭晓

通道

服务商 InfAI ↗

通道R-Kiro

服务商 ByteCatCode ↗

通道O-Max

官方基准

服务商 Novaris ↗

通道O-Max

服务商 ClaudeZZ ↗

通道R-kiro

返回列表