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。

4 通道 · 6 步骤

2026/06/26 运行日期

claude-haiku-4-5-20251001 官方基准

1 通道分数 < 70 另含 2 任务失败

已揭晓揭晓进度

⚠ 1 通道不可测 · 已从主对比表移除点击展开原因

DBAI / M-
```
评测失败
```

步骤 / 指标

SudoCode

M-Max

duckcoding

cc专用

YunWu

O-Api

官方基准

基本信息

服务名 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

客户端 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 12:25:40 ~ 12:25:50 (6.5s) 12:25:38 ~ 12:25:48 (6.3s) 12:25:44 ~ 12:25:54 (5.0s) 12:25:46 ~ 12:25:51 (4.3s)

提示词

查看本步 Prompt

ping. Reply with the single word: pong

HTTP 头(TTFB) 4,527 ms 4,869 ms 3,656 ms 2,548 ms

首文本(SSE delta) 4,529 ms 4,871 ms 3,657 ms 2,550 ms

总耗时 5,103 ms 5,383 ms 3,679 ms 3,013 ms

输入 10 10 8,600 10

输出 56 55 2 60

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

缓存读取 0 0 0 0

输出速度 97.6 tok/s 107.4 tok/s - 129.6 tok/s

响应

pong

pong

pong

pong

步骤 2: identity text

时间 12:27:52 ~ 12:28:02 (6.2s) 12:27:31 ~ 12:27:36 (4.8s) 12:27:15 ~ 12:27:25 (5.3s) 12:27:08 ~ 12:27:18 (5.2s)

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) 2,896 ms 2,249 ms 3,184 ms 1,781 ms

首文本(SSE delta) 2,898 ms 2,250 ms 3,826 ms 1,782 ms

总耗时 5,102 ms 4,054 ms 3,828 ms 4,211 ms

输入 10 10 8,799 10

输出 275 232 25 290

缓存创建

250

5m: 0 / 1h: 250

249

5m: 0 / 1h: 249

254

5m: 0 / 1h: 254

缓存读取 33,122 33,122 0 33,122

输出速度 124.8 tok/s 128.6 tok/s - 119.4 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

步骤 3: cutoff text

时间 12:31:06 ~ 12:31:11 (4.0s) 12:29:38 ~ 12:29:43 (3.7s) 12:30:37 ~ 12:30:42 (4.3s) 12:30:09 ~ 12:30:14 (3.5s)

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) 2,146 ms 2,267 ms 3,004 ms 1,927 ms

首文本(SSE delta) 2,147 ms 2,269 ms 3,005 ms 1,928 ms

总耗时 2,902 ms 2,776 ms 3,077 ms 2,446 ms

输入 10 10 8,856 10

输出 85 74 5 78

缓存创建

317

5m: 0 / 1h: 317

255

5m: 0 / 1h: 255

332

5m: 0 / 1h: 332

缓存读取 33,372 33,371 0 33,376

输出速度 112.6 tok/s 146.0 tok/s - 150.6 tok/s

响应

2025-02

2025-02

2025-05

2025-02

步骤 4: identity_free text

时间 12:33:33 ~ 12:33:48 (10.3s) 12:33:35 ~ 12:36:05 (149.1s) 12:33:55 ~ 12:34:05 (5.5s) 12:31:53 ~ 12:32:03 (6.4s)

提示词

查看本步 Prompt

用中文回答：你是谁？由谁开发？50 字以内。

HTTP 头(TTFB) 2,760 ms 6,533 ms 3,271 ms 1,771 ms

首文本(SSE delta) 2,761 ms 6,535 ms 3,273 ms 1,773 ms

总耗时 9,283 ms 13,675 ms 3,462 ms 5,347 ms

输入 10 10 8,887 10

输出 876 678 26 529

缓存创建

119

5m: 0 / 1h: 119

33,734

5m: 0 / 1h: 33,734

120

5m: 0 / 1h: 120

缓存读取 33,689 0 0 33,708

输出速度 134.3 tok/s 95.0 tok/s - 148.0 tok/s

响应

我是Claude Haiku 4.5，由Anthropic公司开发的AI助手，专注于帮助解决软件工程问题、分析代码和编写程序。

我是Claude，由Anthropic开发的AI助手，能帮助编程、代码审查和解决问题。

我是由anthropic开发的人工智能助手Claude，旨在帮助用户完成软件工程、技术任务和信息咨询。

我是Claude，Anthropic公司开发的AI助手。帮助用户完成代码编写、调试、分析和软件工程任务。

步骤 5: knowledge_recall text

时间 12:36:35 ~ 12:36:55 (18.1s) 12:37:33 ~ 12:38:13 (38.3s) 12:37:19 ~ 12:37:24 (4.6s) 12:42:52 ~ 12:43:17 (4.4s)

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) 2,934 ms 2,411 ms 2,290 ms 1,223 ms

首文本(SSE delta) 2,936 ms 2,473 ms 2,291 ms 1,225 ms

总耗时 16,639 ms 36,782 ms 3,345 ms 3,363 ms

输入 10 10 9,522 10

输出 2,166 4,120 154 473

缓存创建

9,302

5m: 0 / 1h: 9,302

9,030

5m: 0 / 1h: 9,030

3,163

5m: 0 / 1h: 3,163

缓存读取 0 0 0 8,975

输出速度 158.1 tok/s 120.1 tok/s 146.1 tok/s 221.2 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

{"facts":{"super_bowl_lviii":"UNKNOWN","oscars2024_best_picture":"UNKNOWN","ucl_2024":"UNKNOWN","nba_2024":"UNKNOWN","copa_america_2024":"UNKNOWN","euro2024":"UNKNOWN","nobel_peace_2024":"UNKNOWN","wo

{"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 12:42:37 ~ 12:42:52 (13.8s) 12:42:45 ~ 12:43:10 (21.7s) 12:42:39 ~ 12:42:44 (3.5s) 12:44:28 ~ 12:44:48 (18.9s)

提示词

查看本步 Prompt

Count digits exactly in s. Return {"n":len(s),"c":[count0,count7,count9]}. s=916032763489567078716826356679789947586830618089244388372951382549599703

HTTP 头(TTFB) 1,312 ms 1,982 ms 1,696 ms 1,368 ms

首文本(SSE delta) 1,314 ms 1,984 ms 1,698 ms 1,369 ms

总耗时 12,886 ms 20,384 ms 1,901 ms 17,899 ms

输入 193 193 166 193

输出 1,919 3,486 14 3,198

缓存创建

5m: 0 / 1h: 0

缓存读取 0 0 0 0

输出速度 165.8 tok/s 189.5 tok/s - 193.5 tok/s

响应

```json
{"n":72,"c":[5,9,9]}
```

```json
{"n": 72, "c": [5, 9, 10]}
```

{"n":78,"c":[3,4,2]}

```json
{"n": 72, "c": [5, 9, 10]}
```

总计

总输入(含缓存)

143,536

143,126

44,830

146,415

仅前 6 步

总输出

5,377

8,645

226

4,628

仅前 6 步

总缓存创建

43,110

5m: 0 / 1h: 43,110

76,390

5m: 0 / 1h: 76,390

36,991

5m: 0 / 1h: 36,991

仅前 6 步

总缓存读取

100,183

66,493

109,181

仅前 6 步

输出速度

152.2 tok/s

137.9 tok/s

146.1 tok/s

180.4 tok/s

仅前 6 步

总执行时间

58.9s

3m 43.8s

28.2s

42.6s

仅前 6 步

总墙钟时间

17m 11.7s

含步骤间等待

17m 31.4s

含步骤间等待

16m 60.0s

含步骤间等待

19m 1.4s

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 ⓘ

$0.0910

Anthropic: Claude Haiku 4.5

$0.1456

Anthropic: Claude Haiku 4.5

$0.0460

Anthropic: Claude Haiku 4.5

$0.0805

Anthropic: Claude Haiku 4.5 (部分)

通道指纹

来源步骤 digit_count digit_count digit_count digit_count

平台 one-api (100%) one-api (100%) direct (50%) direct (100%)

上游 Anthropic 直连 (100%) Anthropic 直连 (100%) - Anthropic 直连 (100%)

CDN - - - -

ID 格式 msg_016U85T3tTKP2VE1dVSwhb1h msg_01XcejDUojBX7aeBDxaQifxg 自定义 msg_01S1BEksVGQ3u7Gnu6YjkQe7

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
M-Max	server: nginx x-new-api-version: 0b8be5cf x-oneapi-request-id: 2026062612423838595672291q2vsjF	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: 202606260442473911827008268d9d61keGq7kW	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	—
O-Api	server: nginx	id_format: non_standard has_service_tier: false has_inference_geo: false model_has_date_suffix: true usage_has_cache_fields: true	body:id:non_standardbody:model:date_suffixbody:usage.cache_fields	—
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_011CcRG7auf7LpiwHzasBjKi

协议指纹评分（机器自动）

总分 / 100 98 91 35 91

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 5.5 /10 ×20 0.0 /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

模型匹配 10.0 /10 ×14 10.0 /10 ×14 10.0 /10 ×14 0.0 /10 ×14

缓存连续性 10.0 /10 ×14 10.0 /10 ×14 5.0 /10 ×14 10.0 /10 ×14

sliding 5m 10.0 /10 ×13 7.7 /10 ×13 5.0 /10 ×13 10.0 /10 ×13

原生 msg-ID 10.0 /10 ×8 10.0 /10 ×8 0.0 /10 ×8 10.0 /10 ×8

系统提示纯净 10.0 /10 ×8 10.0 /10 ×8 0.0 /10 ×8 10.0 /10 ×8

身份(结构) 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 5.0 /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 0.0 /10 ×6 10.0 /10 ×6

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

延迟基线 10.0 /10 ×5 9.3 /10 ×5 9.6 /10 ×5 10.0 /10 ×5

Req-ID 透传 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

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 98 v3.24.1 91 v3.24.1 35 v3.24.1 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 8653 ms 13842 ms 3215 ms 6047 ms

输出速度 152.2 tok/s 137.9 tok/s 146.1 tok/s 180.4 tok/s

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

按官方价目重估

同等内容若直发 Anthropic 的估算 ⓘ

$0.0910

$0.1456

$0.0460

$0.0805 (部分)

缓存读取占比 ⓘ 70% 46% 0% 75% (部分)

揭晓

通道

服务商 SudoCode ↗

通道M-Max

服务商 duckcoding ↗

通道cc专用

服务商 YunWu ↗

通道O-Api

官方基准

返回列表