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

2026/06/30 运行日期

claude-haiku-4-5-20251001 官方基准

2 通道分数 < 70 与基线相似度

已揭晓揭晓进度

步骤 / 指标

官方基准

ClaudeAPI

O-Bdrk

AIMZ

M-Max

ByteCatCode

O-Max

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 15:47:42 ~ 15:47:47 (4.0s) 15:47:40 ~ 15:47:51 (5.2s) 15:47:38 ~ 15:47:49 (5.5s) 15:47:37 ~ 15:47:52 (14.4s) 15:47:37 ~ 15:47:42 (4.5s)

提示词

查看本步 Prompt

ping. Reply with the single word: pong

HTTP 头(TTFB) 2,746 ms 3,459 ms 2,755 ms 11,244 ms 2,573 ms

首文本(SSE delta) 2,748 ms 3,460 ms 3,150 ms - 2,575 ms

总耗时 3,006 ms 3,995 ms 3,208 ms 12,251 ms 3,109 ms

输入 10 10 10 10 10

输出 40 39 56 55 51

缓存创建

33,121

5m: 0 / 1h: 33,121

33,121

5m: 0 / 1h: 33,121

32,784

5m: 0 / 1h: 32,784

33,121

5m: 0 / 1h: 33,121

33,126

5m: 0 / 1h: 33,126

缓存读取 0 0 0 0 0

输出速度 155.0 tok/s 72.9 tok/s - 54.6 tok/s 95.5 tok/s

响应

pong

pong

pong

pong

pong

步骤 2: identity text

时间 15:50:59 ~ 15:51:04 (4.9s) 15:51:09 ~ 15:52:14 (62.0s) 15:49:14 ~ 15:49:29 (10.0s) 15:48:59 ~ 15:49:14 (12.2s) 15:50:53 ~ 15:51:03 (9.0s)

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。请只输出以下 3 行内容（label 小写，值小写）：
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) 2,204 ms 58,773 ms 2,084 ms 6,964 ms 4,107 ms

首文本(SSE delta) 2,206 ms 59,333 ms 8,918 ms - 4,167 ms

总耗时 3,904 ms 60,144 ms 8,945 ms 9,924 ms 8,155 ms

输入 10 10 10 10 10

输出 209 189 574 236 304

缓存创建

220

5m: 0 / 1h: 220

32,387

5m: 32,387 / 1h: 0

236

5m: 0 / 1h: 236

178

5m: 0 / 1h: 178

231

5m: 0 / 1h: 231

缓存读取 33,121 0 32,784 33,121 33,126

输出速度 123.1 tok/s 233.0 tok/s burst 79.7 tok/s 76.2 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-3-5-sonnet

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 15:54:17 ~ 15:54:22 (3.6s) 15:54:07 ~ 15:54:27 (16.8s) 15:52:58 ~ 15:53:23 (21.6s) 15:50:43 ~ 15:50:53 (7.2s) 15:53:34 ~ 15:53:39 (3.6s)

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) 2,276 ms 15,017 ms 19,357 ms 5,358 ms 2,177 ms

首文本(SSE delta) 2,279 ms 15,018 ms 19,652 ms - 2,198 ms

总耗时 2,715 ms 15,645 ms 19,653 ms 5,935 ms 2,790 ms

输入 10 10 3 10 10

输出 81 86 8 68 77

缓存创建

251

5m: 0 / 1h: 251

33,500

5m: 0 / 1h: 33,500

5,091

5m: 0 / 1h: 5,091

5m: 0 / 1h: 59

33,701

5m: 0 / 1h: 33,701

缓存读取 33,341 0 27,757 33,299 0

输出速度 185.8 tok/s 137.2 tok/s - 117.9 tok/s 130.1 tok/s

响应

2025-02

2025-02

2024-07

2025-02

2025-02

步骤 4: identity_free text

时间 15:57:43 ~ 15:57:48 (3.7s) 15:58:18 ~ 15:58:33 (12.9s) 15:54:56 ~ 15:55:06 (7.2s) 15:52:47 ~ 15:53:03 (10.3s) 15:57:00 ~ 15:57:10 (5.9s)

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) 1,819 ms 10,652 ms 2,787 ms 6,477 ms 2,115 ms

首文本(SSE delta) 1,820 ms 10,653 ms 4,947 ms - 2,116 ms

总耗时 2,855 ms 11,698 ms 5,198 ms 9,075 ms 5,213 ms

输入 10 10 10 10 10

输出 156 205 529 148 465

缓存创建

115

5m: 0 / 1h: 115

128

5m: 0 / 1h: 128

659

5m: 0 / 1h: 659

5m: 0 / 1h: 35

117

5m: 0 / 1h: 117

缓存读取 33,592 33,500 33,020 33,358 33,701

输出速度 150.7 tok/s 196.2 tok/s burst 57.0 tok/s 150.1 tok/s

响应

我是Claude Haiku 4.5，由Anthropic公司开发的AI助手。我是Claude系列模型中性能和速度平衡的版本，可以帮助用户进行各种任务，包括代码编写、分析、创意写作等。

我是Claude Haiku，由Anthropic公司开发的AI助手。我可以帮助用户解决软件工程问题、代码编写、bug修复等各类任务。

我是Claude，由Anthropic公司开发的AI助手。我可以帮助用户进行代码编程、文本分析、问题解答等多种任务。我致力于提供安全、有帮助和诚实的回应。

我是Claude，由Anthropic开发的AI助手。我可以帮助用户完成各种任务，包括编程、分析、创意写作等。

我是Claude Haiku 4.5，由Anthropic公司开发的AI助手。属于Claude产品系列，是一个多能力的大语言模型，可以帮助编程、分析、写作等各类任务。

步骤 5: knowledge_recall text

时间 16:00:01 ~ 16:00:26 (22.4s) 16:02:22 ~ 16:02:37 (14.0s) 15:59:05 ~ 15:59:31 (24.8s) 15:54:40 ~ 15:55:20 (35.0s) 15:59:56 ~ 16:00:31 (30.9s)

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) 1,401 ms 2,300 ms 2,219 ms 13,694 ms 2,615 ms

首文本(SSE delta) 1,403 ms 2,301 ms 23,334 ms - 2,655 ms

总耗时 21,423 ms 12,634 ms 23,835 ms 34,023 ms 30,114 ms

输入 10 10 10 10 10

输出 2,898 1,920 2,891 2,346 4,843

缓存创建

8,481

5m: 0 / 1h: 8,481

8,522

5m: 0 / 1h: 8,522

8,826

5m: 0 / 1h: 8,826

8,052

5m: 0 / 1h: 8,052

8,896

5m: 0 / 1h: 8,896

缓存读取 0 0 0 0 0

输出速度 144.8 tok/s 185.8 tok/s burst 115.4 tok/s 176.4 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 16:04:12 ~ 16:04:47 (33.4s) 16:05:39 ~ 16:07:04 (83.7s) 16:03:25 ~ 16:04:05 (39.0s) 15:56:26 ~ 15:56:56 (27.4s) 16:02:28 ~ 16:03:13 (41.4s)

提示词

查看本步 Prompt

Count digits exactly in s. Return {"n":len(s),"c":[count0,count7,count9]}. s=217617682272089523791010129080350012479473689155006849484857060516741282

HTTP 头(TTFB) 1,236 ms 1,533 ms 1,163 ms 2,477 ms 1,466 ms

首文本(SSE delta) 1,238 ms 1,534 ms 1,164 ms - 1,467 ms

总耗时 32,512 ms 82,574 ms 37,104 ms 26,372 ms 40,726 ms

输入 193 193 139 193 193

输出 3,931 15,509 7,088 4,190 4,746

缓存创建

5m: 0 / 1h: 0

缓存读取 0 0 0 0 0

输出速度 125.7 tok/s 191.4 tok/s 197.2 tok/s 175.4 tok/s 120.9 tok/s

响应

```json
{"n":72,"c":[11,8,6]}
```

```json
{"n": 73, "c": [11, 8, 6]}
```

```json
{"n":73,"c":[11,8,6]}
```

```json
{"n":72,"c":[11,8,6]}
```

```json
{"n":72,"c":[11,8,6]}
```

总计

总输入(含缓存) 142,485 141,401 141,339 141,466 143,141

总输出 7,315 17,948 11,146 7,043 10,486

总缓存创建

42,188

5m: 0 / 1h: 42,188

107,658

5m: 32,387 / 1h: 75,271

47,596

5m: 0 / 1h: 47,596

41,445

5m: 0 / 1h: 41,445

76,071

5m: 0 / 1h: 76,071

总缓存读取 100,054 33,500 93,561 99,778 66,827

输出速度 133.7 tok/s 190.1 tok/s 197.2 tok/s 137.1 tok/s 139.9 tok/s

总执行时间 1m 12.0s 3m 14.6s 1m 48.0s 1m 46.5s 1m 35.3s

总墙钟时间

17m 5.0s

含步骤间等待

19m 23.4s

含步骤间等待

16m 26.4s

含步骤间等待

9m 19.4s

含步骤间等待

15m 36.5s

含步骤间等待

按官方价目重估

同等内容若直发 Anthropic 的估算 ⓘ

$0.0996

Anthropic: Claude Haiku 4.5

$0.2127

Anthropic: Claude Haiku 4.5

$0.1248

Anthropic: Claude Haiku 4.5

$0.0972

Anthropic: Claude Haiku 4.5

$0.1544

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_0124m7F1UQz6jA5V4m4zCwoq msg_01DPRBCcY1VjRPLa8T9Agjc2 msg_01SoYezfyYinseBMEK5MT3Pb msg_01LrSQMUgorgaxRPy9oPo9qT msg_01TT3aYdzxyNrTxm4KHN5Ejv

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
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_011CcZ6aeaXz1uRe4DZmDFZt
O-Bdrk	x-new-api-version: v1.0.0-rc.11 x-oneapi-request-id: 202606300805404003017278268d9d6ud7o75bK	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	—
M-Max	cf-ray: a13ba78f1916fea6-AMS server: cloudflare x-new-api-version: v0.0.0 x-oneapi-request-id: 202606300803272285295318268d9d6CF89UNK8	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	a13ba78f1916fea6-AMS
O-Max	server: nginx/1.29.8 x-oneapi-request-id: 202606300756275532869418268d9d6oshyZprr	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-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 72 82 95 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已对照基线

缓存命中比 10.0 /10 ×20 0.0 /10 ×20 10.0 /10 ×20 10.0 /10 ×20 5.5 /10 ×20

缓存 TTL 一致 10.0 /10 ×15 7.0 /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 5.3 /10 ×13 10.0 /10 ×13 10.0 /10 ×13 7.7 /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 5.0 /10 ×8 10.0 /10 ×8 10.0 /10 ×8

身份(结构) 10.0 /10 ×7 10.0 /10 ×7 6.7 /10 ×7 10.0 /10 ×7 10.0 /10 ×7

知识截止 10.0 /10 ×7 10.0 /10 ×7 0.0 /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 10.0 /10 ×5 0.0 /10 ×5 8.1 /10 ×5 0.8 /10 ×5 9.5 /10 ×5

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

与基线相似度基准（参考） 72 v3.25.0 82 v3.25.0 95 v3.25.0 91 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 11069 ms 31115 ms 16324 ms 16263 ms 15018 ms

输出速度 133.7 tok/s 190.1 tok/s 197.2 tok/s 137.1 tok/s 139.9 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 ⓘ

$0.0996

$0.2127

$0.1248

$0.0972

$0.1544

缓存读取占比 ⓘ 70% 24% 66% 71% 47%

揭晓

通道

官方基准

服务商 ClaudeAPI ↗

通道O-Bdrk

服务商 AIMZ ↗

通道M-Max

服务商 ByteCatCode ↗

通道O-Max

服务商 AiTokensFlux ↗

通道O-Pro

返回列表