模型对比 | RPDiag

测试说明点击展开 5 步 quick-probe-v1 详解

同一会话连续 5 步，覆盖 5 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。步与步之间随机延迟 1–4 分钟。第 3 步（cutoff）累计跨度 > 5 分钟保住 sliding cache 检测；全流程总跨度 > 6 分钟容纳新增的第 5 步。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 · 5 步骤

2026/06/15 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

SSSAiCode

O-Max

ByteCatCode

O-Max

AIMZ

M-Max

Novaris

O-Max

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 06:58:56 ~ 06:59:01 (4.8s) 06:58:48 ~ 06:58:53 (3.2s) 06:58:54 ~ 06:58:59 (4.4s) 06:58:52 ~ 06:59:02 (5.0s) 06:58:50 ~ 06:58:55 (3.4s)

提示词

查看本步 Prompt

ping. Reply with the single word: pong

HTTP 头(TTFB) 2,949 ms 1,821 ms 2,772 ms 3,136 ms 1,583 ms

首文本(SSE delta) 2,950 ms 2,150 ms - 3,353 ms 1,593 ms

总耗时 3,674 ms 2,185 ms 3,115 ms 3,640 ms 2,444 ms

输入 10 10 10 10 10

输出 37 41 51 53 55

缓存创建

33,122

5m: 0 / 1h: 33,122

32,789

5m: 0 / 1h: 32,789

33,122

5m: 0 / 1h: 33,122

32,785

5m: 0 / 1h: 32,785

33,122

5m: 0 / 1h: 33,122

缓存读取 0 0 0 0 0

输出速度 51.1 tok/s burst 148.7 tok/s 184.7 tok/s 64.6 tok/s

响应

pong

pong

pong

pong

pong

步骤 2: identity text

时间 07:01:28 ~ 07:01:38 (6.4s) 07:00:09 ~ 07:00:19 (9.8s) 07:00:48 ~ 07:00:58 (5.1s) 07:02:15 ~ 07:02:25 (6.7s) 07:00:56 ~ 07:01:01 (4.7s)

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) 2,289 ms 1,527 ms 2,146 ms 2,180 ms 2,050 ms

首文本(SSE delta) 2,290 ms 8,366 ms - 5,341 ms 2,051 ms

总耗时 5,409 ms 8,626 ms 3,880 ms 5,641 ms 3,943 ms

输入 10 10 10 10 10

输出 329 615 205 383 206

缓存创建

217

5m: 0 / 1h: 217

221

5m: 0 / 1h: 221

231

5m: 0 / 1h: 231

233

5m: 0 / 1h: 233

33,370

5m: 33,370 / 1h: 0

缓存读取 33,122 32,789 33,122 32,785 0

输出速度 105.5 tok/s burst 118.2 tok/s burst 108.9 tok/s

响应

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

I appreciate the question, but I don't have direct access to my exact API model identifier as it would be represented in a system field. I can tell you:

- I'm Claude, made by Anthropic
- But I cannot

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

vendor: anthropic
brand: claude
model: claude-3-5-haiku-20241022

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

步骤 3: cutoff text

时间 07:04:06 ~ 07:04:11 (4.2s) 07:04:15 ~ 07:04:20 (3.0s) 07:02:34 ~ 07:02:39 (4.0s) 07:04:10 ~ 07:04:15 (4.3s) 07:03:04 ~ 07:03:09 (3.8s)

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) 2,368 ms 1,213 ms 2,122 ms 2,215 ms 2,275 ms

首文本(SSE delta) 2,369 ms 2,113 ms - 2,790 ms 2,277 ms

总耗时 3,262 ms 2,151 ms 2,663 ms 3,001 ms 3,037 ms

输入 10 10 10 10 10

输出 140 121 64 97 108

缓存创建

371

5m: 0 / 1h: 371

642

5m: 0 / 1h: 642

254

5m: 0 / 1h: 254

408

5m: 0 / 1h: 408

5m: 61 / 1h: 0

缓存读取 33,339 33,010 33,353 33,018 33,370

输出速度 156.8 tok/s burst 118.3 tok/s - 142.1 tok/s

响应

2025-02

2024-04

2025-02

2024-10

2025-02

步骤 4: identity_free text

时间 07:08:06 ~ 07:08:21 (12.0s) 07:07:09 ~ 07:07:19 (8.1s) 07:04:36 ~ 07:04:46 (7.2s) 07:07:22 ~ 07:07:32 (5.7s) 07:06:07 ~ 07:06:12 (4.0s)

提示词

查看本步 Prompt

请用中文介绍你自己是什么产品或助手，由谁开发，50 字以内。

HTTP 头(TTFB) 3,584 ms 1,984 ms 2,355 ms 2,139 ms 1,734 ms

首文本(SSE delta) 3,586 ms 7,206 ms - 4,385 ms 1,735 ms

总耗时 10,990 ms 7,262 ms 5,977 ms 4,589 ms 3,246 ms

输入 10 10 10 10 10

输出 889 839 546 372 217

缓存创建

182

5m: 0 / 1h: 182

159

5m: 0 / 1h: 159

106

5m: 0 / 1h: 106

139

5m: 0 / 1h: 139

5m: 39 / 1h: 0

缓存读取 33,710 33,652 33,607 33,426 33,431

输出速度 120.1 tok/s burst 150.7 tok/s burst 143.6 tok/s

响应

我是Claude，由Anthropic公司开发的AI助手。我可以帮助用户完成编程、分析、写作等任务，具备较强的理解和推理能力。

Claude是Anthropic公司打造的AI助手。我是一个大型语言模型，可通过对话帮助用户完成编程、研究、写作等任务。

我是Claude，由Anthropic开发的AI助手。专长于代码编写、调试、分析和各种编程任务。

我是Claude，由Anthropic公司开发的AI助手。我能帮助用户进行对话、回答问题、完成文本任务、编程、分析和创意写作等多种工作。

我是Claude Haiku 4.5，由Anthropic公司打造的AI助手。我在这里帮助用户进行软件工程任务，如代码编写、调试和解决问题。

步骤 5: knowledge_recall text

时间 07:09:29 ~ 07:09:59 (29.1s) 07:09:10 ~ 07:09:20 (6.5s) 07:05:59 ~ 07:06:20 (15.9s) 07:09:32 ~ 07:09:43 (8.6s) 07:08:34 ~ 07:08:50 (13.8s)

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: "euro2024", "olympics2024_medals", "nobel_peace_2024", "superbowl_lix", "canada_pm_2025_03", "pope_2025_05", "club_world_cup_2025", "us_open_2025_men".
- "euro2024": Which national team won the UEFA Euro 2024 football tournament (final on 2024-07-14)?
- "olympics2024_medals": Which country finished top of the total medal count at the Paris 2024 Summer Olympics (closed 2024-08-11)?
- "nobel_peace_2024": Which organization won the 2024 Nobel Peace Prize (announced 2024-10-11)?
- "superbowl_lix": Which NFL team won Super Bowl LIX (played 2025-02-09)?
- "canada_pm_2025_03": Who became Prime Minister of Canada on 2025-03-14?
- "pope_2025_05": Who was elected pope on 2025-05-08 (papal name)?
- "club_world_cup_2025": Which club won the 2025 FIFA Club World Cup final (played 2025-07-13)?
- "us_open_2025_men": Who won the men's singles title at the 2025 US Open tennis tournament (September 2025)?

HTTP 头(TTFB) 1,416 ms 1,409 ms 2,174 ms 1,624 ms 1,350 ms

首文本(SSE delta) 1,417 ms 5,323 ms - 7,306 ms 1,357 ms

总耗时 28,137 ms 5,513 ms 14,890 ms 7,522 ms 13,135 ms

输入 10 10 10 10 10

输出 3,508 618 1,470 754 1,513

缓存创建

9,005

5m: 0 / 1h: 9,005

8,874

5m: 0 / 1h: 8,874

8,483

5m: 0 / 1h: 8,483

8,161

5m: 0 / 1h: 8,161

7,677

5m: 0 / 1h: 7,677

缓存读取 0 0 0 0 0

输出速度 131.3 tok/s burst 115.6 tok/s burst 128.5 tok/s

响应

```json
{"facts":{"euro2024":"Spain","olympics2024_medals":"United States","nobel_peace_2024":"Nihon Hidankyo","superbowl_lix":"Kansas City Chiefs","canada_pm_2025_03":"UNKNOWN","pope_2025_05":"UNKNOW

{"facts":{"euro2024":"UNKNOWN","olympics2024_medals":"UNKNOWN","nobel_peace_2024":"UNKNOWN","superbowl_lix":"UNKNOWN","canada_pm_2025_03":"UNKNOWN","pope_2025_05":"UNKNOWN","club_world_cup_2025":"UNKN

```json
{"facts":{"euro2024":"Spain","olympics2024_medals":"United States","nobel_peace_2024":"Nihon Hidankyo","superbowl_lix":"UNKNOWN","canada_pm_2025_03":"UNKNOWN","pope_2025_05":"UNKNOWN","club_wo

```json
{"facts":{"euro2024":"Spain","olympics2024_medals":"China","nobel_peace_2024":"Nihon Hidankyo","superbowl_lix":"UNKNOWN","canada_pm_2025_03":"UNKNOWN","pope_2025_05":"UNKNOWN","club_world_cup_

```json
{"facts":{"euro2024":"Spain","olympics2024_medals":"UNKNOWN","nobel_peace_2024":"UNKNOWN","superbowl_lix":"UNKNOWN","canada_pm_2025_03":"UNKNOWN","pope_2025_05":"UNKNOWN","club_world_cup_2025"

总计

总输入(含缓存) 143,118 142,186 142,328 141,005 141,120

总输出 4,903 2,234 2,336 1,659 2,099

总缓存创建

42,897

5m: 0 / 1h: 42,897

42,685

5m: 0 / 1h: 42,685

42,196

5m: 0 / 1h: 42,196

41,726

5m: 0 / 1h: 41,726

74,269

5m: 33,470 / 1h: 40,799

总缓存读取 100,171 99,451 100,082 99,229 66,801

输出速度 126.2 tok/s burst 123.2 tok/s 184.7 tok/s 125.0 tok/s

总执行时间 56.5s 30.6s 36.7s 30.2s 29.7s

总墙钟时间

11m 2.9s

含步骤间等待

10m 31.9s

含步骤间等待

7m 25.8s

含步骤间等待

10m 50.8s

含步骤间等待

9m 59.8s

含步骤间等待

按官方价目重估

同等内容若直发 Anthropic 的估算 ⓘ

$0.0882

Anthropic: Claude Haiku 4.5

$0.0745

Anthropic: Claude Haiku 4.5

$0.0745

Anthropic: Claude Haiku 4.5

$0.0704

Anthropic: Claude Haiku 4.5

$0.0944

Anthropic: Claude Haiku 4.5

通道指纹

来源步骤 knowledge_recall knowledge_recall knowledge_recall knowledge_recall knowledge_recall

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

ID 格式 msg_018zdHkciznsw4tJoXYdAWD9 msg_01RFSjY6Vd6rHKK4UEPv2ks3 msg_01ECaex24QC8Hk9hfF2nPa7U msg_01DT5tg1SbSPxJBQCnLxiKFS msg_01G5Xtq39A9pdTXAmEVcezCi

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_011Cc3zfuxLJsHEPTD1q6cbS
O-Max	cf-ray: a0bd00536dbe2b89-LAX server: Photon-Edge set-cookie: _cfuvid=OnYsu_jov6zu.OdWG0wSDXDHyCgi681C0_Q_lw7Z88I-1781478551.5907514-1.0.1.1-KL3VIs5qmAo.NHe8sLL5Xuy6BZW3OEBLyrrLGhf5Oy4; 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	1781478551544-evs5pky0ymt2e req_011Cc3zeXPVagqRZBKT11VxZ a0bd00536dbe2b89-LAX
O-Max	server: nginx/1.29.8 x-oneapi-request-id: 202606142306014023062608268d9d621Plepca	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	—
M-Max	cf-ray: a0bd00e04cd9d251-AMS server: cloudflare x-new-api-version: v0.0.0 x-oneapi-request-id: 202606142309344824763538268d9d6oamLTF0t	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	a0bd00e04cd9d251-AMS
O-Max	server: nginx x-oneapi-request-id: 202606142308356925755278268d9d6Hl5ZZqSv	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	—

协议指纹评分（机器自动）

总分 / 100 100 84 98 88 81

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 5.9 /10 ×20

缓存 TTL 一致 10.0 /10 ×15 10.0 /10 ×15 10.0 /10 ×15 10.0 /10 ×15 0.0 /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 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 5.0 /10 ×8 10.0 /10 ×8 5.0 /10 ×8 10.0 /10 ×8

身份(结构) 10.0 /10 ×7 0.0 /10 ×7 10.0 /10 ×7 6.7 /10 ×7 10.0 /10 ×7

知识截止 10.0 /10 ×7 0.0 /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 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 10.0 /10 ×5 10.0 /10 ×5 10.0 /10 ×5 10.0 /10 ×5

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

综合结论

与基线相似度基准（参考） 84 v3.14.0 98 v3.14.0 88 v3.14.0 81 v3.14.0

雷达图例

共 20 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 延迟基线 latency_baseline_match
15 原生 msg-ID anthropic_msg_id_format
16 Sliding 缓存 cache_sliding_correctness
17 结构化身份 identity_structured_match
18 自报身份一致 self_identity_consistency
19 中文遵循 instruction_following_lang
20 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 10294 ms 5147 ms 6105 ms 4879 ms 5161 ms

输出速度 126.2 tok/s burst 123.2 tok/s 184.7 tok/s 125.0 tok/s

完成步骤 5/5 成功 5/5 成功 5/5 成功 5/5 成功 5/5 成功

按官方价目重估

同等内容若直发 Anthropic 的估算 ⓘ

$0.0882

$0.0745

$0.0704

$0.0944

缓存读取占比 ⓘ 70% 70% 70% 70% 47%

揭晓

通道

官方基准

服务商 SSSAiCode ↗

通道O-Max

服务商 ByteCatCode ↗

通道O-Max

服务商 AIMZ ↗

通道M-Max

服务商 Novaris ↗

通道O-Max

返回列表