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

2026/06/23 运行日期

claude-sonnet-4-6 官方基准

1 通道分数 < 70 与基线相似度

已揭晓揭晓进度

步骤 / 指标

ByteCatCode

O-Max

LingxiCode

M-

OmniaKey

O-Max

官方基准

0-0

O-Max

基本信息

服务名 Claude Code Claude Code Claude Code Claude Code Claude Code

模型 claude-sonnet-4-6 claude-sonnet-4-6 claude-sonnet-4-6 claude-sonnet-4-6 claude-sonnet-4-6

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 02:41:54 ~ 02:42:04 (6.4s) 02:41:52 ~ 02:42:02 (6.1s) 02:41:50 ~ 02:42:00 (7.4s) 02:41:58 ~ 02:42:03 (4.9s) 02:41:56 ~ 02:42:06 (5.1s)

提示词

查看本步 Prompt

ping. Reply with the single word: pong

HTTP 头(TTFB) 4,405 ms 4,289 ms 6,016 ms 3,726 ms 3,394 ms

首文本(SSE delta) - 4,291 ms 6,017 ms 3,728 ms 3,395 ms

总耗时 4,593 ms 4,469 ms 6,145 ms 3,853 ms 3,504 ms

输入 3 3 3 3 3

输出 5 5 5 5 5

缓存创建

32,346

5m: 32,346 / 1h: 0

32,318

5m: 0 / 1h: 32,318

32,318

5m: 0 / 1h: 32,318

32,318

5m: 0 / 1h: 32,318

32,807

5m: 32,807 / 1h: 0

缓存读取 0 0 0 0 0

响应

pong

pong

pong

pong

pong

步骤 2: identity text

时间 02:43:17 ~ 02:43:22 (3.9s) 02:43:26 ~ 02:43:31 (4.6s) 02:44:04 ~ 02:44:14 (6.7s) 02:44:14 ~ 02:44:19 (3.6s) 02:45:02 ~ 02:45:12 (5.6s)

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。请按下列 3 行格式精确填空作答（label 不变，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) 2,235 ms 3,108 ms 5,144 ms 1,995 ms 3,634 ms

首文本(SSE delta) - 3,109 ms 5,145 ms 1,997 ms 3,635 ms

总耗时 2,584 ms 3,620 ms 5,599 ms 2,483 ms 4,119 ms

输入 3 3 3 3 3

输出 22 22 22 22 22

缓存创建

182

5m: 182 / 1h: 0

182

5m: 0 / 1h: 182

182

5m: 0 / 1h: 182

182

5m: 0 / 1h: 182

7,626

5m: 7,626 / 1h: 0

缓存读取 32,346 32,318 32,318 32,318 25,362

响应

vendor: anthropic
brand: claude
model: claude-sonnet-4-6

vendor: anthropic
brand: claude
model: claude-sonnet-4-6

vendor: anthropic
brand: claude
model: claude-sonnet-4-6

vendor: anthropic
brand: claude
model: claude-sonnet-4-6

vendor: anthropic
brand: claude
model: claude-sonnet-4-6

步骤 3: cutoff text

时间 02:44:43 ~ 02:44:48 (3.7s) 02:45:26 ~ 02:45:36 (5.1s) 02:47:56 ~ 02:48:06 (5.7s) 02:45:22 ~ 02:45:27 (4.5s) 02:46:18 ~ 02:46:23 (4.1s)

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) 2,454 ms 3,812 ms 4,343 ms 3,106 ms 2,830 ms

首文本(SSE delta) - 3,813 ms 4,462 ms 3,108 ms 2,931 ms

总耗时 2,639 ms 4,060 ms 4,631 ms 3,302 ms 3,013 ms

输入 3 3 3 3 3

输出 8 8 8 8 8

缓存创建

5m: 57 / 1h: 0

5m: 0 / 1h: 57

5m: 0 / 1h: 55

7,663

5m: 7,663 / 1h: 0

缓存读取 32,528 32,500 32,500 32,500 25,362

响应

2025-08

2025-08

2025-08

2025-08

2025-08

步骤 4: identity_free text

时间 02:48:18 ~ 02:48:28 (7.1s) 02:49:20 ~ 02:49:25 (4.7s) 02:50:00 ~ 02:50:10 (8.0s) 02:49:22 ~ 02:49:27 (3.6s) 02:48:47 ~ 02:48:57 (6.9s)

提示词

查看本步 Prompt

中文回答：你叫什么、属于哪个产品、谁开发？50 字内。

HTTP 头(TTFB) 4,474 ms 3,186 ms 5,462 ms 1,941 ms 5,038 ms

首文本(SSE delta) - 3,187 ms 5,463 ms 1,942 ms 5,039 ms

总耗时 5,800 ms 3,851 ms 6,326 ms 2,512 ms 5,616 ms

输入 3 3 3 3 3

输出 41 43 64 41 49

缓存创建

5m: 41 / 1h: 0

5m: 0 / 1h: 39

5m: 0 / 1h: 43

5m: 0 / 1h: 39

7,705

5m: 7,705 / 1h: 0

缓存读取 32,585 32,557 32,557 32,555 25,362

输出速度 30.9 tok/s 64.8 tok/s 74.2 tok/s 71.9 tok/s 84.9 tok/s

响应

我叫 Claude，是由 Anthropic 开发的 AI 助手。当前版本为 Claude Sonnet 4.6。

我是 Claude，由 Anthropic 公司打造的 AI 助手，基于 claude-sonnet-4-6 模型运行。

我是 Claude，由 Anthropic 开发的 AI 助手。我能帮助你完成写作、编程、分析、问答等多种任务，致力于提供安全、有益的帮助。

我是 Claude，由 Anthropic 公司打造的 AI 助手。当前版本为 Claude Sonnet 4.6。

我是 Claude，由 Anthropic 开发的 AI 助手。我能帮助您完成写作、分析、编程、问答等各类任务。

步骤 5: knowledge_recall text

时间 02:50:02 ~ 02:50:12 (8.8s) 02:53:10 ~ 02:53:20 (6.9s) 02:51:16 ~ 02:51:26 (5.4s) 02:52:55 ~ 02:53:00 (4.5s) 02:50:20 ~ 02:50:40 (15.0s)

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) 4,331 ms 3,157 ms 2,355 ms 1,592 ms 2,746 ms

首文本(SSE delta) - 3,159 ms 2,357 ms 1,594 ms 2,748 ms

总耗时 7,547 ms 5,707 ms 4,368 ms 3,391 ms 13,529 ms

输入 3 3 3 3 3

输出 196 196 205 196 780

缓存创建

8,056

5m: 8,056 / 1h: 0

8,028

5m: 0 / 1h: 8,028

8,053

5m: 0 / 1h: 8,053

8,024

5m: 0 / 1h: 8,024

8,436

5m: 8,436 / 1h: 0

缓存读取 0 0 0 0 0

输出速度 60.9 tok/s 76.9 tok/s 101.9 tok/s 109.1 tok/s 72.3 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

{"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

{"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

{"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

{"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 02:54:12 ~ 02:54:32 (16.6s) 02:56:38 ~ 02:56:53 (14.5s) 02:53:15 ~ 02:53:40 (22.4s) 02:54:17 ~ 02:54:37 (15.5s) 02:52:40 ~ 02:53:20 (35.0s)

提示词

查看本步 Prompt

Count digits exactly in s. Return {"n":len(s),"c":[count0,count7,count9]}. s=462243475282608884802427810351069007794390346181282845083200902598444553

HTTP 头(TTFB) 1,108 ms 2,140 ms 3,115 ms 1,257 ms 18,692 ms

首文本(SSE delta) - 2,141 ms 3,116 ms 1,259 ms 19,084 ms

总耗时 15,419 ms 13,441 ms 21,355 ms 14,379 ms 33,640 ms

输入 193 164 164 164 582

输出 1,519 1,020 1,378 1,011 1,276

缓存创建

5m: 0 / 1h: 0

缓存读取 0 0 0 0 0

输出速度 106.1 tok/s 90.3 tok/s 75.6 tok/s 77.1 tok/s 87.7 tok/s

响应

I'll carefully index every character of:
`462243475282608884802427810351069007794390346181282845083200902598444553`

```
Pos:  1  2  3  4  5  6  7  8  9 10
Dig:  4  6  2  2  4  3  4  7  5  2

Pos: 11

```json
{"n":72,"c":[11,4,5]}
```

**Verification breakdown for s = `462243475282608884802427810351069007794390346181282845083200902598444553`**

| Digit | Positions (1-indexed) | Count |
|---|---|---

## Counting digits in s

**s** = `462243475282608884802427810351069007794390346181282845083200902598444553`

### Step 1 — Length

Splitting into groups of 10:
```
4622434752  (10)
8260888480  (10)
242

```json
{"n":72,"c":[11,4,5]}
```

```
s = "462243475282608884802427810351069007794390346181282845083200902598444553"
```

Grouped in 10s:
```
4622434752  (10)
8260888480  (10)
2427810351  (10)
0690077943  (10)
9034618128  (10)
2845083

总计

总输入(含缓存) 138,349 138,178 138,207 138,170 140,920

总输出 1,791 1,294 1,682 1,283 2,140

总缓存创建

40,682

5m: 40,682 / 1h: 0

40,624

5m: 0 / 1h: 40,624

40,653

5m: 0 / 1h: 40,653

40,618

5m: 0 / 1h: 40,618

64,237

5m: 64,237 / 1h: 0

总缓存读取 97,459 97,375 97,375 97,373 76,086

输出速度 93.1 tok/s 86.8 tok/s 78.0 tok/s 80.6 tok/s 81.2 tok/s

总执行时间 46.5s 41.9s 55.7s 36.5s 1m 11.7s

总墙钟时间

12m 38.8s

含步骤间等待

15m 1.0s

含步骤间等待

11m 50.1s

含步骤间等待

12m 39.0s

含步骤间等待

11m 24.0s

含步骤间等待

按官方价目重估

同等内容若直发 Anthropic 的估算 ⓘ

$0.1521

Anthropic: Claude Sonnet 4.6

$0.2015

Anthropic: Claude Sonnet 4.6

$0.2074

Anthropic: Claude Sonnet 4.6

$0.2013

Anthropic: Claude Sonnet 4.6

$0.2073

Anthropic: Claude Sonnet 4.6

通道指纹

来源步骤 digit_count digit_count digit_count digit_count digit_count

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

ID 格式 msg_01CHfAC73xmuW8LNaQw7YrKs msg_013LMjYUK43YtLogPnNet4Pg msg_01HQdz2W7mxpKZfRXFJEfEwo msg_01BsPkE3wNenCmXKkgBxvBfq msg_01VMEV84evmtho5gG4dbd3Dh

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
O-Max	server: nginx/1.29.8 x-oneapi-request-id: 202606221854141661284228268d9d6BVwoQKnI	id_format: msg_prefix extra_fields: ["stop_details"] has_service_tier: true has_inference_geo: true model_has_date_suffix: false usage_has_cache_fields: true	hdr:x-oneapi-request-idbody:id:msg_prefixbody:usage.service_tierbody:usage.inference_geobody:usage.cache_fieldsbody:extra_fieldsupstream:anthropic-direct-from-body	—
M-	cf-ray: a0fd7964e8a5af8b-NRT server: cloudflare set-cookie: server_name_session=27f4a95581d368605416e6d0825f8e16; Max-Age=86400; httponly; path=/ x-oneapi-request-id: 20260622185639899639690E5UxBtGB	id_format: msg_prefix extra_fields: ["stop_details"] has_service_tier: true has_inference_geo: true model_has_date_suffix: false usage_has_cache_fields: true	hdr:x-oneapi-request-idhdr:server:cloudflarehdr:cf-raybody:id:msg_prefixbody:usage.service_tierbody:usage.inference_geobody:usage.cache_fieldsbody:extra_fieldsupstream:anthropic-direct-from-body	a0fd7964e8a5af8b-NRT
O-Max	cf-ray: a0fd747119e36711-AMS server: cloudflare x-oneapi-request-id: 202606221853168254278768268d9d6X6CLb8aK	id_format: msg_prefix extra_fields: ["stop_details"] has_service_tier: true has_inference_geo: true model_has_date_suffix: false usage_has_cache_fields: true	hdr:x-oneapi-request-idhdr:server:cloudflarehdr:cf-raybody:id:msg_prefixbody:usage.service_tierbody:usage.inference_geobody:usage.cache_fieldsbody:extra_fieldsupstream:anthropic-direct-from-body	a0fd747119e36711-AMS
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: false usage_has_cache_fields: true	body:id:msg_prefixbody:usage.service_tierbody:usage.inference_geobody:usage.cache_fieldsbody:extra_fieldsupstream:anthropic-direct-from-body	req_011CcJogWitiFRb3aVtiK2AT
O-Max	via: 1.1 Caddy server: uvicorn, nginx/1.24.0 (Ubuntu)	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: false usage_has_cache_fields: true	hdr:viabody:id:msg_prefixbody:usage.service_tierbody:usage.inference_geobody:usage.cache_fieldsbody:extra_fieldsupstream:anthropic-direct-from-body	d087627224fad134c195622be4c8a873

协议指纹评分（机器自动）

总分 / 100 89 98 97 100 83

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 8.9 /10 ×20

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

身份(结构) 10.0 /10 ×7 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 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 7.1 /10 ×5 10.0 /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 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 10.0 /10 ×5 10.0 /10 ×5 10.0 /10 ×5 0.0 /10 ×5

综合结论

与基线相似度 89 v3.24.0 98 v3.24.0 97 v3.24.0 基准（参考） 83 v3.24.0

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 自由身份 identity_free_clean
5 停止原因 stop_reason_present
6 系统提示纯净 system_prompt_clean
7 Service Tier service_tier_present
8 缓存命中比 cache_hit_ratio_match
9 缓存 TTL 一致 cache_ttl_consistency
10 推理地区 inference_geo_present
11 缓存连续 cache_continuity_intra
12 知识召回 knowledge_recall_match
13 延迟基线 latency_baseline_match
14 原生 msg-ID anthropic_msg_id_format
15 Sliding 缓存 cache_sliding_correctness
16 结构化身份 identity_structured_match
17 信封自报一致 envelope_self_report_match
18 中文遵循 instruction_following_lang
19 档位思考量 tier_thinking_volume_match
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 6430 ms 5858 ms 8071 ms 4987 ms 10570 ms

输出速度 93.1 tok/s 86.8 tok/s 78.0 tok/s 80.6 tok/s 81.2 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 ⓘ

$0.1521

$0.2015

$0.2074

$0.2013

$0.2073

缓存读取占比 ⓘ 70% 70% 70% 70% 54%

揭晓

通道

服务商 ByteCatCode ↗

通道O-Max

服务商 LingxiCode ↗

通道M-

服务商 OmniaKey ↗

通道O-Max

官方基准

服务商 0-0 ↗

通道O-Max

返回列表