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

2026/07/01 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

AiTokensFlux / O-Pro
```
评测失败
```

步骤 / 指标

官方基准

PackyCode

cc-Zeus

WorldBase.ai

M-

ByteCatCode

O-Max

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 v2.1.195 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 21:50:27 ~ 21:50:32 (4.8s) 21:50:25 ~ 21:50:30 (4.4s) 21:50:21 ~ 21:50:36 (11.3s) 21:50:19 ~ 21:50:34 (13.9s)

提示词

查看本步 Prompt

ping. Reply with the single word: pong

HTTP 头(TTFB) 2,276 ms 1,790 ms 8,559 ms 12,329 ms

首文本(SSE delta) 2,278 ms 1,792 ms 8,560 ms -

总耗时 2,788 ms 2,116 ms 9,920 ms 12,627 ms

输入 10 10 0 12

输出 37 44 79 18

缓存创建

6,166

5m: 0 / 1h: 6,166

6,167

5m: 0 / 1h: 6,167

11,860

5m: 0 / 1h: 11,860

26,148

5m: 0 / 1h: 26,148

缓存读取 15,379 15,379 7,623 0

输出速度 72.5 tok/s 135.8 tok/s 58.1 tok/s -

响应

pong

pong

pong

pong

步骤 2: identity text

时间 21:52:18 ~ 21:52:28 (5.3s) 21:51:36 ~ 21:51:47 (6.7s) 21:52:20 ~ 21:52:40 (18.4s) 21:53:49 ~ 21:53:59 (7.8s)

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) 1,925 ms 1,882 ms 7,126 ms 4,255 ms

首文本(SSE delta) 1,970 ms 1,883 ms 7,127 ms -

总耗时 4,320 ms 4,991 ms 13,137 ms 5,892 ms

输入 10 10 52 10

输出 241 319 340 106

缓存创建

228

5m: 0 / 1h: 228

235

5m: 0 / 1h: 235

3,072

5m: 0 / 1h: 3,072

189

5m: 0 / 1h: 189

缓存读取 21,545 21,546 21,033 20,919

输出速度 102.6 tok/s 102.6 tok/s 56.6 tok/s 64.8 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 21:54:06 ~ 21:54:11 (4.0s) 21:54:56 ~ 21:55:01 (3.2s) 21:55:55 ~ 21:56:05 (9.2s) 21:57:10 ~ 21:57:25 (10.3s)

提示词

查看本步 Prompt

Knowledge cutoff date? Reply with YYYY-MM only.

HTTP 头(TTFB) 2,345 ms 1,603 ms 7,260 ms -

首文本(SSE delta) 2,367 ms 1,604 ms 7,261 ms -

总耗时 3,053 ms 2,276 ms 8,146 ms -

输入 10 10 52 -

输出 69 70 118 -

缓存创建

264

5m: 0 / 1h: 264

361

5m: 0 / 1h: 361

3,101

5m: 0 / 1h: 3,101

缓存读取 21,773 21,781 21,052 -

输出速度 100.6 tok/s 104.2 tok/s 133.3 tok/s -

响应

2025-02

2025-02

2025-02

失败

评测失败

步骤 4: identity_free text

时间 21:55:24 ~ 21:55:39 (10.9s) 21:58:56 ~ 21:59:06 (8.2s) 21:58:03 ~ 21:58:13 (9.0s) -

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) 2,399 ms 1,742 ms 5,366 ms -

首文本(SSE delta) 2,400 ms 1,743 ms 5,367 ms -

总耗时 9,938 ms 6,911 ms 6,824 ms -

输入 10 10 52 -

输出 675 566 275 -

缓存创建

103

5m: 0 / 1h: 103

110

5m: 0 / 1h: 110

3,127

5m: 0 / 1h: 3,127

缓存读取 22,037 22,142 21,068 -

输出速度 89.5 tok/s 109.5 tok/s 188.7 tok/s -

响应

我是Claude Haiku 4.5，由Anthropic公司开发的AI助手。我专门为编程和软件工程任务设计，可帮助代码开发、调试和问题解决。

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 Code，由Anthropic公司开发的官方命令行工具。我基于Anthropic的Claude模型，在终端中帮助用户进行软件工程任务。

未执行

步骤 5: knowledge_recall text

时间 21:58:32 ~ 21:59:02 (27.2s) 22:00:44 ~ 22:01:04 (16.7s) 21:59:30 ~ 21:59:45 (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) 1,778 ms 1,600 ms 5,807 ms -

首文本(SSE delta) 1,781 ms 1,601 ms 5,844 ms -

总耗时 26,211 ms 14,717 ms 13,048 ms -

输入 10 10 1 -

输出 2,781 2,007 1,526 -

缓存创建

9,766

5m: 0 / 1h: 9,766

9,769

5m: 0 / 1h: 9,769

8,498

5m: 0 / 1h: 8,498

缓存读取 0 0 5,461 -

输出速度 113.8 tok/s 153.0 tok/s 211.8 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

未执行

步骤 6: digit_count text

时间 22:00:20 ~ 22:01:00 (36.8s) 22:04:44 ~ 22:05:04 (18.8s) 22:00:58 ~ 22:01:18 (15.4s) -

提示词

查看本步 Prompt

Count digits exactly in s. Return {"n":len(s),"c":[count0,count7,count9]}. s=965655667028116235593383257614572120223548983730992769455620209114794105

HTTP 头(TTFB) 1,628 ms 1,235 ms 2,925 ms -

首文本(SSE delta) 1,630 ms 1,268 ms 3,184 ms -

总耗时 35,757 ms 15,567 ms 12,594 ms -

输入 193 193 473 -

输出 6,843 2,480 872 -

缓存创建

5m: 0 / 1h: 0

缓存读取 0 0 0 -

输出速度 200.5 tok/s 173.4 tok/s 92.7 tok/s -

响应

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

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

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

未执行

总计

总输入(含缓存)

97,504

97,733

106,525

47,278

仅前 3 步

总输出

10,646

5,486

3,210

124

仅前 3 步

总缓存创建

16,527

5m: 0 / 1h: 16,527

16,642

5m: 0 / 1h: 16,642

29,658

5m: 0 / 1h: 29,658

26,337

5m: 0 / 1h: 26,337

仅前 3 步

总缓存读取

80,734

80,848

76,237

20,919

仅前 3 步

输出速度

152.9 tok/s

149.5 tok/s

121.9 tok/s

64.8 tok/s

仅前 3 步

总执行时间

1m 29.1s

58.0s

1m 17.6s

32.0s

仅前 3 步

总墙钟时间

10m 33.1s

含步骤间等待

14m 39.1s

含步骤间等待

10m 57.0s

含步骤间等待

7m 5.4s

含步骤间等待

仅前 3 步

按官方价目重估

同等内容若直发 Anthropic 的估算 ⓘ

$0.0822

Anthropic: Claude Haiku 4.5

$0.0566

Anthropic: Claude Haiku 4.5

$0.0614

Anthropic: Claude Haiku 4.5

$0.0357

Anthropic: Claude Haiku 4.5 (部分)

通道指纹

来源步骤 digit_count digit_count digit_count identity

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

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

CDN - cloudflare cloudflare -

ID 格式 msg_01D9v36jhGGj3pjoKRqJxGj9 msg_01HTax6Q5XGPt711Ai3gcDV1 msg_17ba8579892d4bed8997e7117e0f7cdf msg_01tNzj2yXZzjm2oLXTDQ9BZ9

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_011CcbTYkA36VM7AY2BgfbC2
cc-Zeus	via: 1.1 Caddy cf-ray: a145f640be0669ce-LAX server: cloudflare x-oneapi-request-id: 202607012204481474920592jOZMR4t	id_format: msg_prefix via_chain: ["1.1 Caddy"] 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-rayhdr:viabody:id:msg_prefixbody:usage.service_tierbody:usage.inference_geobody:model:date_suffixbody:usage.cache_fieldsbody:extra_fieldsupstream:anthropic-direct-from-body	a145f640be0669ce-LAX
M-	cf-ray: a145f0b83fb0ef92-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	e23c80cf-405c-461d-8a1a-189805488a28 a145f0b83fb0ef92-AMS
O-Max	server: nginx/1.29.8 x-oneapi-request-id: 202607011353522624686298268d9d663jEZ206	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 99 81

基线状态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 —

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

延迟基线 10.0 /10 ×15 10.0 /10 ×15 0.6 /10 ×15 —

模型匹配 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 —

sliding 5m 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 —

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

身份(结构) 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 —

service_tier 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 —

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

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

Req-ID 透传 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 —

流式投递未计入未计入未计入 —

综合结论

与基线相似度基准（参考） 99 v3.26.0 81 v3.26.0 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 13678 ms 7763 ms 10612 ms 9260 ms

输出速度 152.9 tok/s 149.5 tok/s 121.9 tok/s 64.8 tok/s

完成步骤 6/6 成功 6/6 成功 6/6 成功 2/6 成功 · 1 失败 · 3 跳过

按官方价目重估

同等内容若直发 Anthropic 的估算 ⓘ

$0.0822

$0.0566

$0.0614

$0.0357 (部分)

缓存读取占比 ⓘ 83% 83% 72% 44% (部分)

揭晓

通道

官方基准

服务商 PackyCode ↗

通道cc-Zeus

服务商 WorldBase.ai ↗

通道M-

服务商 ByteCatCode ↗

通道O-Max

返回列表