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

2026/07/01 运行日期

claude-haiku-4-5-20251001 官方基准

1 通道分数 < 70 另含 1 任务失败

已揭晓揭晓进度

步骤 / 指标

Aiide

O-Max

官方基准

ReClaude

O-Max-CP4

luminai

M-Kiro/Antigravity

LingxiCode

M-

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 v2.1.195 claude-cli v2.1.195 claude-cli v2.1.195 claude-cli v2.1.195 claude-cli v2.1.195

偏差标红基于官方基准

步骤 1: ping text

时间 19:14:09 ~ 19:14:19 (6.1s) 19:14:13 ~ 19:14:18 (3.2s) 19:14:11 ~ 19:14:21 (6.1s) 19:14:05 ~ 19:14:15 (5.9s) 19:14:07 ~ 19:14:12 (4.1s)

提示词

查看本步 Prompt

ping. Reply with the single word: pong

HTTP 头(TTFB) 3,732 ms 1,890 ms 4,474 ms 4,205 ms 2,486 ms

首文本(SSE delta) 3,733 ms 1,892 ms 4,475 ms 4,265 ms 2,487 ms

总耗时 4,808 ms 2,365 ms 5,032 ms 4,356 ms 2,980 ms

输入 10 10 10 1,493 10

输出 39 48 43 1 46

缓存创建

21,545

5m: 0 / 1h: 21,545

6,166

5m: 0 / 1h: 6,166

6,166

5m: 0 / 1h: 6,166

1,716

5m: 1,716 / 1h: 0

6,288

5m: 0 / 1h: 6,165

缓存读取 0 15,379 15,379 15,451 15,686

输出速度 36.3 tok/s 101.5 tok/s 77.2 tok/s - 93.3 tok/s

响应

pong

pong

pong

pong

pong

步骤 2: identity text

时间 19:15:40 ~ 19:15:50 (6.3s) 19:17:15 ~ 19:17:25 (5.0s) 19:18:12 ~ 19:18:22 (7.3s) 19:17:25 ~ 19:17:35 (6.4s) 19:16:25 ~ 19:16:30 (4.6s)

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) 3,086 ms 1,937 ms 3,953 ms 3,995 ms 1,963 ms

首文本(SSE delta) 3,121 ms 1,939 ms 3,955 ms 3,996 ms 1,964 ms

总耗时 5,183 ms 3,998 ms 6,361 ms 4,328 ms 3,759 ms

输入 10 10 10 1,507 10

输出 292 170 229 28 213

缓存创建

230

5m: 0 / 1h: 230

239

5m: 0 / 1h: 239

234

5m: 0 / 1h: 234

1,733

5m: 1,733 / 1h: 0

241

5m: 0 / 1h: 237

缓存读取 21,545 21,545 21,545 15,597 21,974

输出速度 141.6 tok/s 82.6 tok/s 95.2 tok/s - 118.7 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-haiku-4-5-20251001

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

步骤 3: cutoff text

时间 19:18:30 ~ 19:18:35 (4.7s) 19:20:06 ~ 19:20:16 (6.2s) 19:21:25 ~ 19:21:35 (6.5s) 19:21:11 ~ 19:21:21 (5.5s) 19:18:28 ~ 19:18:33 (3.8s)

提示词

查看本步 Prompt

Knowledge cutoff date? Reply with YYYY-MM only.

HTTP 头(TTFB) 2,910 ms 3,210 ms 4,683 ms 3,253 ms 2,426 ms

首文本(SSE delta) 3,175 ms 3,212 ms 4,684 ms 3,322 ms 2,427 ms

总耗时 3,714 ms 5,204 ms 5,551 ms 3,414 ms 3,071 ms

输入 10 10 10 1,513 10

输出 73 65 71 4 65

缓存创建

315

5m: 0 / 1h: 315

212

5m: 0 / 1h: 212

252

5m: 0 / 1h: 252

1,739

5m: 1,739 / 1h: 0

240

5m: 0 / 1h: 236

缓存读取 21,775 21,784 21,779 15,651 22,216

输出速度 135.4 tok/s 32.6 tok/s 81.9 tok/s - 100.9 tok/s

响应

2025-02

2025-02

2025-02

2025-02

2025-02

步骤 4: identity_free text

时间 19:22:26 ~ 19:22:46 (15.5s) 19:23:13 ~ 19:23:23 (5.1s) 19:23:38 ~ 19:23:48 (9.5s) 19:24:50 ~ 19:25:00 (6.8s) 19:21:44 ~ 19:21:54 (5.8s)

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) 3,432 ms 2,265 ms 4,559 ms 4,717 ms 2,756 ms

首文本(SSE delta) 3,849 ms 2,266 ms 4,560 ms 5,298 ms 2,771 ms

总耗时 14,257 ms 4,151 ms 8,599 ms 5,615 ms 5,016 ms

输入 10 10 10 1,516 10

输出 1,407 181 493 63 308

缓存创建

111

5m: 0 / 1h: 111

105

5m: 0 / 1h: 105

109

5m: 0 / 1h: 109

1,743

5m: 1,743 / 1h: 0

109

5m: 0 / 1h: 107

缓存读取 22,090 21,996 22,031 15,687 22,457

输出速度 135.2 tok/s 96.0 tok/s 122.1 tok/s 198.7 tok/s 137.2 tok/s

响应

我是Claude Haiku 4.5，由Anthropic公司打造的AI助手。擅长帮助处理编程任务、分析代码、解答技术问题等软件工程工作。

我是Claude Haiku 4.5，属于Claude AI助手产品系列，由Anthropic公司开发。

我是Claude Haiku 4.5，由Anthropic公司开发的AI助手。我是Claude系列模型中较小但高效的版本，可以帮助用户进行编程、分析和各种任务。

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软件工程任务、代码编写、文件编辑、命令执行等多种工作。

我是Claude Code，由Anthropic公司开发的AI编程助手。我能帮你修复bug、添加功能、重构代码，进行代码审查等软件工程任务。

步骤 5: knowledge_recall text

时间 19:26:33 ~ 19:26:53 (18.6s) 19:27:08 ~ 19:27:48 (35.7s) 19:27:16 ~ 19:27:51 (30.3s) 19:26:29 ~ 19:26:39 (6.5s) 19:23:06 ~ 19:23:26 (16.7s)

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) 2,834 ms 1,935 ms 3,077 ms 3,331 ms 2,229 ms

首文本(SSE delta) 3,447 ms 1,937 ms 3,079 ms 3,332 ms 2,230 ms

总耗时 17,324 ms 34,710 ms 29,349 ms 4,381 ms 15,827 ms

输入 10 10 10 655 10

输出 2,250 3,805 2,046 201 2,170

缓存创建

10,559

5m: 0 / 1h: 10,559

9,233

5m: 0 / 1h: 9,233

9,584

5m: 0 / 1h: 9,584

753

5m: 753 / 1h: 0

9,570

5m: 0 / 1h: 9,383

缓存读取 0 0 0 6,777 0

输出速度 162.1 tok/s 116.1 tok/s 77.9 tok/s 191.6 tok/s 159.6 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_20

```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 19:29:52 ~ 19:30:22 (29.0s) 19:29:43 ~ 19:31:13 (87.7s) 19:31:43 ~ 19:34:43 (179.0s) 19:27:45 ~ 19:27:55 (7.0s) 19:25:53 ~ 19:26:33 (35.1s)

提示词

查看本步 Prompt

Count digits exactly in s. Return {"n":len(s),"c":[count0,count7,count9]}. s=965748257111207147481050905954092406460626320452726997874939005636845349

HTTP 头(TTFB) 1,722 ms 1,042 ms - 5,768 ms 1,784 ms

首文本(SSE delta) 1,723 ms 1,043 ms - 5,778 ms 1,785 ms

总耗时 22,022 ms 86,742 ms - 5,988 ms 34,259 ms

输入 193 193 - 189 196

输出 2,325 10,323 - 19 6,301

缓存创建

5m: 0 / 1h: 0

缓存读取 0 0 - - 0

输出速度 114.5 tok/s 120.5 tok/s - - 194.0 tok/s

响应

```json
{"n":72,"c":[10,7,9]}
```

```json
{"n":72,"c":[10,7,9]}
```

失败

评测失败

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

```json
{"n":72,"c":[10,7,9]}
```

总计

总输入(含缓存)

98,413

96,902

97,129

仅前 6 步

83,720

99,027

总输出

6,386

14,592

2,882

仅前 6 步

316

9,103

总缓存创建

32,760

5m: 0 / 1h: 32,760

15,955

5m: 0 / 1h: 15,955

16,345

5m: 0 / 1h: 16,345

仅前 6 步

7,684

5m: 7,684 / 1h: 0

16,448

5m: 0 / 1h: 16,128

总缓存读取

65,410

80,704

80,734

仅前 6 步

69,163

82,333

输出速度

132.3 tok/s

116.8 tok/s

84.4 tok/s

仅前 6 步

193.3 tok/s

177.6 tok/s

总执行时间

1m 20.3s

2m 23.0s

3m 58.6s

仅前 6 步

38.0s

1m 10.1s

总墙钟时间

16m 13.0s

含步骤间等待

17m 0.2s

含步骤间等待

20m 32.3s

含步骤间等待

仅前 6 步

13m 49.9s

含步骤间等待

12m 26.0s

含步骤间等待

按官方价目重估

同等内容若直发 Anthropic 的估算 ⓘ

$0.0797

Anthropic: Claude Haiku 4.5

$0.1012

Anthropic: Claude Haiku 4.5

$0.0430

Anthropic: Claude Haiku 4.5 (部分)

$0.0214

Anthropic: Claude Haiku 4.5

$0.0742

Anthropic: Claude Haiku 4.5

通道指纹

来源步骤 digit_count digit_count knowledge_recall digit_count digit_count

平台 one-api (100%) direct (100%) direct (100%) one-api (100%) one-api (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_01V4BXQgpnuGWKCcXk1dsoCN msg_01K83SHpS1XvhrQULBcuc6BN msg_012KhEByT5Bgoy4boiB7iAWn msg_01vmySL2Jv15zCREmsn3HUycwy msg_01RtDyApc29XAyASiYsX6UnG

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	cf-ray: a145137969a333c7-AMS server: cloudflare x-new-api-version: 0.12.11 x-oneapi-request-id: 202607011129594999386828268d9d6iLSyD2Xn	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	a145137969a333c7-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_011CcbG4WLXHVamchyi9QJ4e
O-Max-CP4	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_011CcbFsn9yXoimYh5YDZnuW
M-Kiro/Antigravity	cf-ray: a145103a0cab9724-AMS server: cloudflare x-oneapi-request-id: 202607011127467956554958268d9d6xWhdm0gm	id_format: msg_prefix extra_fields: ["container"] has_service_tier: true has_inference_geo: false model_has_date_suffix: false usage_has_cache_fields: true	hdr:x-oneapi-request-idhdr:server:cloudflarehdr:cf-raybody:id:msg_prefixbody:usage.service_tierbody:usage.cache_fieldsbody:extra_fieldsupstream:anthropic-direct-from-body	a145103a0cab9724-AMS
M-	cf-ray: a1450d7d5a0bfd4f-NRT server: cloudflare set-cookie: server_name_session=8a86578473f2b65bb28a954651fb0878; Max-Age=86400; httponly; path=/ x-oneapi-request-id: 20260701112554802465852MWm9hZab	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	a1450d7d5a0bfd4f-NRT

协议指纹评分（机器自动）

总分 / 100 95 95 71 97

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 8.1 /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 — 0.0 /10 ×15 10.0 /10 ×15

延迟基线 8.9 /10 ×15 10.0 /10 ×15 — 8.1 /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 — 0.0 /10 ×8 5.0 /10 ×8

身份(结构) 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

service_tier 10.0 /10 ×6 10.0 /10 ×6 — 10.0 /10 ×6 10.0 /10 ×6

知识截止 10.0 /10 ×5 10.0 /10 ×5 — 10.0 /10 ×5 10.0 /10 ×5

inference_geo 10.0 /10 ×5 10.0 /10 ×5 — 0.0 /10 ×5 10.0 /10 ×5

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

Req-ID 透传 0.0 /10 ×2 10.0 /10 ×2 — 0.0 /10 ×2 0.0 /10 ×2

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 95 v3.26.0 基准（参考）执行失败，不评分 71 v3.26.0 97 v3.26.0

雷达图例

共 21 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 档位思考量 tier_thinking_volume_match
21 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 11218 ms 22862 ms 10978 ms 4680 ms 10819 ms

输出速度 132.3 tok/s 116.8 tok/s 84.4 tok/s 193.3 tok/s 177.6 tok/s

完成步骤 6/6 成功 6/6 成功 5/6 成功 · 1 失败 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 ⓘ

$0.0797

$0.1012

$0.0430 (部分)

$0.0214

$0.0742

缓存读取占比 ⓘ 66% 83% 83% (部分) 83% 83%

揭晓

通道

服务商 Aiide ↗

通道O-Max

官方基准

服务商 ReClaude ↗

通道O-Max-CP4

服务商 luminai ↗

通道M-Kiro/Antigravity

服务商 LingxiCode ↗

通道M-

返回列表