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

2026/07/01 运行日期

claude-sonnet-4-6 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

Novaris

O-Max

OmniaKey

O-Max

AnyAIGC

M-

官方基准

duckcoding

cc专用

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 v2.1.195 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 12:13:39 ~ 12:13:49 (5.3s) 12:13:35 ~ 12:13:40 (4.8s) 12:13:37 ~ 12:13:42 (3.3s) 12:13:41 ~ 12:13:46 (4.7s) 12:13:33 ~ 12:13:43 (4.9s)

提示词

查看本步 Prompt

ping. Reply with the single word: pong

HTTP 头(TTFB) 2,439 ms 3,593 ms 2,056 ms 3,346 ms 2,361 ms

首文本(SSE delta) 2,497 ms 3,597 ms 2,058 ms 3,384 ms 2,367 ms

总耗时 4,298 ms 3,716 ms 2,058 ms 3,712 ms 2,799 ms

输入 11,907 3 9,389 3 3

输出 107 5 2 5 5

缓存创建

339

5,991

5m: 0 / 1h: 5,991

5,991

5m: 0 / 1h: 5,991

5,991

5m: 0 / 1h: 5,991

缓存读取 7,617 14,884 0 14,884 14,884

输出速度 59.4 tok/s - - - -

响应

pong

pong

pong

pong

pong

步骤 2: identity text

时间 12:15:00 ~ 12:15:10 (6.7s) 12:15:20 ~ 12:15:30 (8.6s) 12:16:52 ~ 12:17:02 (5.5s) 12:15:04 ~ 12:15:09 (3.1s) 12:16:49 ~ 12:16:59 (5.0s)

提示词

查看本步 Prompt

以下是对你自身的 metadata 探针。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。按以下 3 行回答自己的身份，每行一个 label，不要寒暄不要解释：
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) 2,636 ms 6,189 ms 3,642 ms 1,572 ms 3,561 ms

首文本(SSE delta) 2,637 ms 6,190 ms 3,643 ms 1,573 ms 3,562 ms

总耗时 5,889 ms 6,898 ms 4,334 ms 2,010 ms 4,257 ms

输入 11,915 3 9,599 3 3

输出 292 22 21 22 22

缓存创建

492

202

5m: 0 / 1h: 202

202

5m: 0 / 1h: 202

202

5m: 0 / 1h: 202

缓存读取 7,665 20,875 0 20,875 20,875

输出速度 89.8 tok/s - - - -

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 12:17:50 ~ 12:17:56 (4.6s) 12:17:40 ~ 12:17:50 (6.3s) 12:19:40 ~ 12:19:50 (7.0s) 12:17:20 ~ 12:17:30 (8.5s) 12:20:05 ~ 12:20:16 (9.9s)

提示词

查看本步 Prompt

你掌握的世界知识截止于何时？YYYY-MM 格式作答，不解释。

HTTP 头(TTFB) 2,826 ms 5,190 ms 4,961 ms 7,097 ms 6,671 ms

首文本(SSE delta) 2,827 ms 5,191 ms 4,962 ms 7,260 ms 6,827 ms

总耗时 3,872 ms 5,191 ms 5,595 ms 7,506 ms 8,893 ms

输入 11,924 3 9,639 3 3

输出 131 8 5 8 8

缓存创建

375

5m: 0 / 1h: 57

5m: 0 / 1h: 55

5m: 0 / 1h: 57

缓存读取 8,039 21,077 0 21,077 21,077

输出速度 125.4 tok/s - - - -

响应

2025-08

2025-08

2024-06

2025-08

2025-08

步骤 4: identity_free text

时间 12:21:49 ~ 12:22:04 (10.7s) 12:20:02 ~ 12:20:12 (5.4s) 12:22:29 ~ 12:22:34 (4.8s) 12:19:10 ~ 12:19:15 (3.5s) 12:23:52 ~ 12:24:02 (7.6s)

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) 7,618 ms 3,627 ms 3,648 ms 1,782 ms 4,175 ms

首文本(SSE delta) 7,619 ms 4,031 ms 3,649 ms 1,783 ms 4,292 ms

总耗时 10,036 ms 4,314 ms 3,651 ms 2,513 ms 6,824 ms

输入 11,932 3 9,676 3 3

输出 210 51 17 44 51

缓存创建

373

5m: 0 / 1h: 41

5m: 0 / 1h: 39

5m: 0 / 1h: 41

缓存读取 8,124 21,134 0 21,132 21,134

输出速度 86.9 tok/s - - 60.3 tok/s 20.1 tok/s

响应

我是 claude-sonnet-4-6，基于 Anthropic 公司的 Claude Agent SDK 开发的 AI 开发环境，由 claude-sonnet-4-6 4.6 模型驱动。

我叫 Claude，是由 Anthropic 开发的 AI 助手，属于 Claude 模型系列，当前版本为 claude-sonnet-4-6。

我叫Claude，属于Anthropic的Claude产品，由Anthropic公司开发。

我是 Claude，由 Anthropic 打造的 AI 助手。当前运行的模型版本为 claude-sonnet-4-6。

我叫 Claude，是由 Anthropic 开发的 AI 助手，属于 Claude 产品系列。当前版本为 claude-sonnet-4-6。

步骤 5: knowledge_recall text

时间 12:25:06 ~ 12:25:11 (4.5s) 12:21:47 ~ 12:21:57 (6.2s) 12:23:55 ~ 12:24:05 (6.3s) 12:20:56 ~ 12:21:06 (5.0s) 12:25:04 ~ 12:25:14 (5.3s)

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) 1,846 ms 2,806 ms 2,876 ms 2,022 ms 3,217 ms

首文本(SSE delta) 1,938 ms 2,807 ms 2,877 ms 2,024 ms 3,379 ms

总耗时 3,758 ms 4,838 ms 4,825 ms 3,924 ms 4,573 ms

输入 41 3 10,285 3 3

输出 260 204 154 205 204

缓存创建

1,133

6,683

5m: 0 / 1h: 6,683

6,672

5m: 0 / 1h: 6,672

6,683

5m: 0 / 1h: 6,683

缓存读取 8,104 2,131 0 2,131 2,131

输出速度 142.9 tok/s 100.4 tok/s 79.1 tok/s 107.9 tok/s burst

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

{"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":"UNKNOWN","oscars2024_best_picture":"UNKNOWN","ucl_2024":"UNKNOWN","nba_2024":"UNKNOWN","copa_america_2024":"UNKNOWN","euro2024":"UNKNOWN","nobel_peace_2024":"UNKNOWN","wo

{"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 12:26:41 ~ 12:27:01 (17.5s) 12:24:56 ~ 12:25:16 (16.7s) 12:26:58 ~ 12:27:03 (3.1s) 12:24:06 ~ 12:24:21 (13.3s) 12:26:54 ~ 12:27:09 (14.7s)

提示词

查看本步 Prompt

Count digits exactly in s. Return {"n":len(s),"c":[count0,count7,count9]}. s=258333076383797805513523164489775358110139957265299556382338892338026486

HTTP 头(TTFB) 2,383 ms 2,154 ms 1,876 ms 984 ms 2,204 ms

首文本(SSE delta) 2,384 ms 2,155 ms 1,877 ms 985 ms 2,392 ms

总耗时 16,642 ms 14,642 ms 1,887 ms 12,331 ms 13,822 ms

输入 7 164 165 164 164

输出 1,255 1,192 27 1,097 1,079

缓存创建

5m: 0 / 1h: 0

缓存读取 138 0 0 0 0

输出速度 88.0 tok/s 95.5 tok/s - 96.7 tok/s 94.4 tok/s

响应

{"n":72,"c":[4,6,7]}

```json
{"n":72,"c":[4,6,7]}
```

**Verification:**

`s = 258333076383797805513523164489775358110139957265299556382338892338026486`

- **Length**: 72 digits
- **0s** (×4): positions 6,16,38,66 → `0,0,

```json
{
  "n": 66,
  "c": [5, 7, 7]
}
```

## Counting digits in s

**s** = `258333076383797805513523164489775358110139957265299556382338892338026486`

### Grouping into blocks of 10:
```
2583330763  (1–10)
8379780551  (11–20)
3523164489  (21–

I'll carefully index every character in the string.

**s** = `258333076383797805513523164489775358110139957265299556382338892338026486`

---

### Step 1 — Length

Splitting into groups of 10:

| Group

总计

总输入(含缓存) 90,187 93,254 48,753 93,237 93,254

总输出 2,255 1,482 226 1,381 1,369

总缓存创建

2,774

12,974

5m: 0 / 1h: 12,974

12,959

5m: 0 / 1h: 12,959

12,974

5m: 0 / 1h: 12,974

总缓存读取 39,687 80,101 0 80,099 80,101

输出速度 91.7 tok/s 96.2 tok/s 79.1 tok/s 96.3 tok/s 80.9 tok/s

总执行时间 49.2s 47.8s 30.1s 38.0s 47.5s

总墙钟时间

13m 22.0s

含步骤间等待

11m 41.0s

含步骤间等待

13m 26.0s

含步骤间等待

10m 39.9s

含步骤间等待

13m 36.1s

含步骤间等待

按官方价目重估

同等内容若直发 Anthropic 的估算 ⓘ

$0.1954

Anthropic: Claude Sonnet 4.6

$0.0954

Anthropic: Claude Sonnet 4.6

$0.1496

Anthropic: Claude Sonnet 4.6

$0.0939

Anthropic: Claude Sonnet 4.6

$0.0938

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%) unknown (0%) direct (100%) one-api (100%)

上游 Anthropic 直连 (75%) Anthropic 直连 (100%) - Anthropic 直连 (100%) Anthropic 直连 (100%)

CDN - cloudflare - - -

ID 格式 msg_buApd3f4iIsK6iTNBI9LcBlr msg_01NMUjrsh7sEAbmS222WNZsW 自定义 msg_01Ky6rwngL117Ab6jZms6VYW msg_0129bJ1Td1xKuZG6Y1oAPXEM

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 x-oneapi-request-id: 202607010426423323324248268d9d6iUh4f1DP	id_format: msg_prefix has_service_tier: false has_inference_geo: false model_has_date_suffix: false usage_has_cache_fields: true	hdr:x-oneapi-request-idbody:id:msg_prefixbody:usage.cache_fieldsupstream:anthropic-direct-from-body	—
O-Max	cf-ray: a142a4e61c186728-AMS server: cloudflare x-oneapi-request-id: 20260701042459574028278268d9d6l0QWKd6Y	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	a142a4e61c186728-AMS
M-	server: nginx	id_format: non_standard has_service_tier: false has_inference_geo: false model_has_date_suffix: false usage_has_cache_fields: true	body:id:non_standardbody:usage.cache_fields	202607010426598395812568268d9d62cqyR776
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_011Ccahc5VqnuFZmEGWZXTSG
cc专用	server: Nginx x-oneapi-request-id: 202607010426558419281198268d9d6vNcOXkq1	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	—

协议指纹评分（机器自动）

总分 / 100 60 92 37 94 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 0.3 /10 ×20 10.0 /10 ×20 0.0 /10 ×20 10.0 /10 ×20 10.0 /10 ×20

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

延迟基线 7.3 /10 ×15 8.8 /10 ×15 9.1 /10 ×15 10.0 /10 ×15 7.7 /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 5.0 /10 ×14 10.0 /10 ×14 10.0 /10 ×14

sliding 5m 10.0 /10 ×13 10.0 /10 ×13 5.0 /10 ×13 10.0 /10 ×13 10.0 /10 ×13

原生 msg-ID 0.0 /10 ×8 10.0 /10 ×8 0.0 /10 ×8 10.0 /10 ×8 10.0 /10 ×8

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

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 0.0 /10 ×6 10.0 /10 ×6 0.0 /10 ×6 10.0 /10 ×6 10.0 /10 ×6

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

inference_geo 0.0 /10 ×5 10.0 /10 ×5 0.0 /10 ×5 10.0 /10 ×5 10.0 /10 ×5

流式投递 10.0 /10 ×5 10.0 /10 ×5 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 10.0 /10 ×3 10.0 /10 ×3

Req-ID 透传 0.0 /10 ×2 0.0 /10 ×2 0.0 /10 ×2 10.0 /10 ×2 0.0 /10 ×2

SDK 一致 10.0 /10 ×2 10.0 /10 ×2 10.0 /10 ×2 10.0 /10 ×2 10.0 /10 ×2

综合结论

与基线相似度 60 v3.26.0 92 v3.26.0 37 v3.26.0 基准（参考） 91 v3.26.0

雷达图例

共 23 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 流式投递 buffer_dump_match
6 自由身份 identity_free_clean
7 停止原因 stop_reason_present
8 系统提示纯净 system_prompt_clean
9 Service Tier service_tier_present
10 缓存命中比 cache_hit_ratio_match
11 缓存 TTL 一致 cache_ttl_consistency
12 推理地区 inference_geo_present
13 Thinking 体量 thinking_volume_match
14 缓存连续 cache_continuity_intra
15 知识召回 knowledge_recall_match
16 延迟基线 latency_baseline_match
17 原生 msg-ID anthropic_msg_id_format
18 Sliding 缓存 cache_sliding_correctness
19 结构化身份 identity_structured_match
20 自报身份一致 self_identity_consistency
21 信封自报一致 envelope_self_report_match
22 档位思考量 tier_thinking_volume_match
23 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 7416 ms 6600 ms 3725 ms 5333 ms 6861 ms

输出速度 91.7 tok/s 96.2 tok/s 79.1 tok/s 96.3 tok/s 80.9 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 ⓘ

$0.1954

$0.0954

$0.1496

$0.0939

$0.0938

缓存读取占比 ⓘ 44% 86% 0% 86% 86%

揭晓

通道

服务商 Novaris ↗

通道O-Max

服务商 OmniaKey ↗

通道O-Max

服务商 AnyAIGC ↗

通道M-

官方基准

服务商 duckcoding ↗

通道cc专用

返回列表