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 另含 1 任务失败

已揭晓揭晓进度

步骤 / 指标

官方基准

OmniaKey

O-Max

HongMaCC

u-cc

FoxCode

R-ultra

WorldBase.ai

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 v2.1.160 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 09:22:43 ~ 09:22:48 (4.0s) 09:22:39 ~ 09:22:49 (5.7s) 09:22:41 ~ 09:22:46 (3.9s) 09:22:35 ~ 09:22:40 (3.8s) 09:22:37 ~ 09:22:47 (6.2s)

提示词

查看本步 Prompt

ping. Reply with the single word: pong

HTTP 头(TTFB) 2,555 ms 3,893 ms 2,446 ms 2,137 ms 3,777 ms

首文本(SSE delta) 2,556 ms 3,894 ms 2,447 ms 2,138 ms 4,101 ms

总耗时 2,852 ms 4,352 ms 2,447 ms 2,846 ms 4,276 ms

输入 10 10 0 33,022 5,086

输出 46 50 36 54 49

缓存创建

33,122

5m: 0 / 1h: 33,122

33,122

5m: 0 / 1h: 33,122

5m: 0 / 1h: 0

28,818

5m: 28,818 / 1h: 0

缓存读取 0 0 0 0 0

输出速度 155.4 tok/s 109.2 tok/s burst 76.3 tok/s -

响应

pong

pong

pong

pong

pong

步骤 2: identity text

时间 09:24:43 ~ 09:24:48 (4.7s) 09:24:59 ~ 09:25:09 (8.5s) 09:24:26 ~ 09:24:37 (6.5s) 09:23:57 ~ 09:24:07 (6.6s) 09:26:31 ~ 09:26:51 (16.3s)

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) 1,626 ms 4,228 ms 5,174 ms 3,317 ms -

首文本(SSE delta) 1,627 ms 4,229 ms 5,175 ms 3,318 ms -

总耗时 3,726 ms 7,040 ms 5,175 ms 5,717 ms -

输入 10 10 0 33,589 -

输出 224 205 146 301 -

缓存创建

240

5m: 0 / 1h: 240

244

5m: 0 / 1h: 244

5m: 0 / 1h: 0

缓存读取 33,122 33,122 0 0 -

输出速度 106.7 tok/s 72.9 tok/s burst 125.5 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

失败

评测失败

步骤 3: cutoff text

时间 09:28:39 ~ 09:28:44 (3.5s) 09:26:53 ~ 09:27:03 (5.4s) 09:28:37 ~ 09:28:42 (3.2s) 09:26:01 ~ 09:26:06 (3.9s) -

提示词

查看本步 Prompt

Knowledge cutoff date? Reply with YYYY-MM only.

HTTP 头(TTFB) 2,009 ms 3,478 ms 2,119 ms 2,141 ms -

首文本(SSE delta) 2,012 ms 3,479 ms 2,120 ms 2,142 ms -

总耗时 2,534 ms 4,154 ms 2,120 ms 3,047 ms -

输入 10 10 0 34,269 -

输出 63 62 43 68 -

缓存创建

247

5m: 0 / 1h: 247

228

5m: 0 / 1h: 228

5m: 0 / 1h: 0

缓存读取 33,362 33,366 0 0 -

输出速度 120.7 tok/s 91.9 tok/s burst 75.1 tok/s -

响应

2025-02

2025-02

2025-02

2025-02

未执行

步骤 4: identity_free text

时间 09:32:30 ~ 09:32:40 (7.7s) 09:29:11 ~ 09:29:21 (8.7s) 09:30:05 ~ 09:30:10 (4.1s) 09:27:08 ~ 09:27:18 (6.2s) -

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) 1,575 ms 3,517 ms 2,816 ms 2,094 ms -

首文本(SSE delta) 1,580 ms 3,518 ms 2,817 ms 2,095 ms -

总耗时 6,712 ms 5,884 ms 2,818 ms 5,421 ms -

输入 10 10 0 34,686 -

输出 772 199 81 675 -

缓存创建

101

5m: 0 / 1h: 101

100

5m: 0 / 1h: 100

5m: 0 / 1h: 0

缓存读取 33,609 33,594 0 0 -

输出速度 150.4 tok/s 84.1 tok/s burst 202.9 tok/s -

响应

我是Claude Haiku 4.5，由Anthropic公司打造的AI助手，专门帮助编程和技术任务。

我是Claude Haiku 4.5，由Anthropic公司打造的AI助手。擅长代码编写、分析和交流。知识库截至2025年2月。

我是 Claude，Anthropic 开发的 AI 助手，帮助用户完成软件工程等任务。

我是Claude，由Anthropic公司开发的AI助手，可以帮助您完成各种任务。

未执行

步骤 5: knowledge_recall text

时间 09:33:55 ~ 09:34:15 (17.1s) 09:32:59 ~ 09:33:24 (22.1s) 09:31:48 ~ 09:31:58 (6.5s) 09:30:07 ~ 09:30:22 (14.3s) -

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) 1,527 ms 2,257 ms 5,265 ms 1,649 ms -

首文本(SSE delta) 1,529 ms 2,258 ms 5,266 ms 1,651 ms -

总耗时 16,158 ms 21,000 ms 5,399 ms 13,287 ms -

输入 10 10 0 10,689 -

输出 2,736 2,229 971 1,875 -

缓存创建

9,100

5m: 0 / 1h: 9,100

8,511

5m: 0 / 1h: 8,511

5m: 0 / 1h: 0

缓存读取 0 0 0 0 -

输出速度 187.0 tok/s 118.9 tok/s - 161.1 tok/s -

响应

{"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

```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":"Kansas City Chiefs","oscars2024_best_picture":"Oppenheimer","ucl_2024":"Real Madrid","nba_2024":"Boston Celtics","copa_america_2024":"Argentina","euro2024":"Spain","nobel

```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

未执行

步骤 6: digit_count text

时间 09:36:22 ~ 09:36:52 (25.7s) 09:37:20 ~ 09:37:40 (18.8s) 09:35:22 ~ 09:35:42 (15.3s) 09:31:26 ~ 09:31:41 (10.5s) -

提示词

查看本步 Prompt

Count digits exactly in s. Return {"n":len(s),"c":[count0,count7,count9]}. s=824212051637693030269948275849528027090066410184778910837042863060275012

HTTP 头(TTFB) 1,109 ms 1,608 ms 13,722 ms 1,056 ms -

首文本(SSE delta) 1,110 ms 1,609 ms 13,723 ms 1,057 ms -

总耗时 24,662 ms 17,656 ms 14,048 ms 9,744 ms -

输入 193 193 0 209 -

输出 3,037 1,905 3,008 2,153 -

缓存创建

5m: 0 / 1h: 0

缓存读取 0 0 0 0 -

输出速度 128.9 tok/s 118.7 tok/s burst 247.8 tok/s -

响应

```json
{"n": 72, "c": [13, 7, 6]}
```

```json
{"n":72,"c":[13,7,6]}
```

{"n":72,"c":[13,7,6]}

```json
{"n":72,"c":[13,7,6]}
```

未执行

总计

总输入(含缓存)

143,146

142,530

146,464

33,904

仅前 2 步

总输出

6,878

4,650

4,285

5,126

仅前 2 步

总缓存创建

42,810

5m: 0 / 1h: 42,810

42,205

5m: 0 / 1h: 42,205

5m: 0 / 1h: 0

28,818

5m: 28,818 / 1h: 0

仅前 2 步

总缓存读取

100,093

100,082

仅前 2 步

输出速度

148.8 tok/s

113.1 tok/s

burst

185.3 tok/s

仅前 2 步

总执行时间

1m 2.8s

1m 9.3s

39.4s

45.2s

22.5s

仅前 2 步

总墙钟时间

14m 9.0s

含步骤间等待

15m 1.1s

含步骤间等待

13m 1.1s

含步骤间等待

9m 5.8s

含步骤间等待

4m 13.3s

含步骤间等待

仅前 2 步

按官方价目重估

同等内容若直发 Anthropic 的估算 ⓘ

$0.0982

Anthropic: Claude Haiku 4.5

$0.0863

Anthropic: Claude Haiku 4.5

>=$0.0214

Anthropic: Claude Haiku 4.5 (成本不完整)

$0.1721

Anthropic: Claude Haiku 4.5

$0.0278

Anthropic: Claude Haiku 4.5 (部分)

通道指纹

来源步骤 digit_count digit_count digit_count digit_count ping

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 - cloudflare cloudflare

ID 格式 msg_01HFoH3eF1UwxyJpdBvx4wXb msg_01GwSyXez42fAHMnBVMaEdLz msg_0e7f1878c3a9ed6a64771c61 msg_01mcBnb2jKFcxdGBEG7MV5c1 msg_c9b5778d06384a508c534662c3dd83ee

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_011CcYb1LEgDZ6CCqxfhV7bt
O-Max	cf-ray: a13972011a8efe3c-AMS server: cloudflare x-oneapi-request-id: 202606300137225769228898268d9d6ZuSm7fm0	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	a13972011a8efe3c-AMS
u-cc	server: nginx x-new-api-version: 1.0.17 x-oneapi-request-id: 20260630013524769414128268d9d6FfYxtT5B	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	—
R-ultra	cf-ray: a13969579b1ae041-NRT server: cloudflare x-new-api-version: v0.0.0 x-oneapi-request-id: 202606300131273866318758268d9d6nsxhAmHt	id_format: msg_prefix 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_fieldsupstream:anthropic-direct-from-body	a13969579b1ae041-NRT
M-	cf-ray: a1395c770e4fdf76-AMS server: cloudflare	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: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	bc477523-8bd3-443c-9bb7-7f7ac7fea5f6 a1395c770e4fdf76-AMS

协议指纹评分（机器自动）

总分 / 100 98 95 60 67

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 8.3 /10 ×20 8.4 /10 ×20 0.0 /10 ×20 0.0 /10 ×20 —

缓存 TTL 一致 10.0 /10 ×15 10.0 /10 ×15 0.0 /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 —

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

sliding 5m 10.0 /10 ×13 10.0 /10 ×13 5.0 /10 ×13 5.0 /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 未计入 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 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 —

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

延迟基线 10.0 /10 ×5 8.1 /10 ×5 9.5 /10 ×5 10.0 /10 ×5 —

Req-ID 透传 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 —

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

与基线相似度基准（参考） 95 v3.25.0 60 v3.25.0 67 v3.25.0 执行失败，不评分

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 9441 ms 10014 ms 5335 ms 6677 ms 4276 ms

输出速度 148.8 tok/s 113.1 tok/s burst 185.3 tok/s -

完成步骤 6/6 成功 6/6 成功 6/6 成功 6/6 成功 1/6 成功 · 1 失败 · 4 跳过

按官方价目重估

同等内容若直发 Anthropic 的估算 ⓘ

$0.0982

$0.0863

>=$0.0214 (成本不完整)

$0.1721

$0.0278 (部分)

缓存读取占比 ⓘ 70% 70% - 0% 0% (部分)

揭晓

通道

官方基准

服务商 OmniaKey ↗

通道O-Max

服务商 HongMaCC ↗

通道u-cc

服务商 FoxCode ↗

通道R-ultra

服务商 WorldBase.ai ↗

通道M-

返回列表