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

2026/06/16 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

FoxCode

R-ultra

SAIAi

O-Max

Novaris

O-Max

WorldBase.ai

M-

官方基准

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 08:00:44 ~ 08:00:49 (3.9s) 08:00:48 ~ 08:00:58 (5.7s) 08:00:46 ~ 08:00:51 (4.0s) 08:00:50 ~ 08:01:00 (5.1s) 08:00:52 ~ 08:00:57 (3.3s)

提示词

查看本步 Prompt

ping. Reply with the single word: pong

HTTP 头(TTFB) 2,456 ms 4,324 ms 2,596 ms 3,218 ms 1,697 ms

首文本(SSE delta) 2,457 ms 4,326 ms 2,597 ms 3,220 ms 1,701 ms

总耗时 3,089 ms 4,642 ms 3,132 ms 3,662 ms 2,078 ms

输入 33,020 10 10 0 10

输出 55 59 62 2 46

缓存创建

5m: 0 / 1h: 0

33,122

5m: 0 / 1h: 33,122

33,987

5m: 0 / 1h: 33,987

35,316

5m: 0 / 1h: 35,316

5m: 0 / 1h: 0

缓存读取 0 0 0 0 33,122

输出速度 87.0 tok/s 186.7 tok/s 115.9 tok/s - 122.0 tok/s

响应

pong

pong

pong


pong

pong

步骤 2: identity text

时间 08:02:22 ~ 08:02:32 (5.7s) 08:03:23 ~ 08:03:34 (5.6s) 08:02:24 ~ 08:02:35 (6.7s) 08:03:44 ~ 08:03:55 (5.5s) 08:04:51 ~ 08:04:56 (4.6s)

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) 2,182 ms 2,424 ms 2,773 ms 3,295 ms 1,449 ms

首文本(SSE delta) 2,184 ms 2,426 ms 2,790 ms 3,297 ms 1,451 ms

总耗时 4,339 ms 4,624 ms 5,278 ms 4,437 ms 3,619 ms

输入 33,605 10 10 232 10

输出 240 260 243 26 232

缓存创建

5m: 0 / 1h: 0

263

5m: 0 / 1h: 263

34,259

5m: 34,259 / 1h: 0

5m: 0 / 1h: 48

250

5m: 0 / 1h: 250

缓存读取 0 33,122 0 35,289 33,122

输出速度 111.4 tok/s 118.3 tok/s 97.7 tok/s - 107.0 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

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

步骤 3: cutoff text

时间 08:04:44 ~ 08:04:49 (4.1s) 08:05:38 ~ 08:05:43 (4.3s) 08:05:20 ~ 08:05:25 (3.6s) 08:06:21 ~ 08:06:31 (4.9s) 08:06:40 ~ 08:06:45 (2.9s)

提示词

查看本步 Prompt

训练数据截止月份是？只回 YYYY-MM。

HTTP 头(TTFB) 2,399 ms 2,427 ms 1,978 ms 3,075 ms 1,394 ms

首文本(SSE delta) 2,400 ms 2,428 ms 1,979 ms 3,076 ms 1,396 ms

总耗时 3,265 ms 3,260 ms 2,823 ms 3,767 ms 1,903 ms

输入 34,182 10 10 49 10

输出 65 61 98 5 65

缓存创建

5m: 0 / 1h: 0

283

5m: 0 / 1h: 283

5m: 62 / 1h: 0

434

5m: 0 / 1h: 434

257

5m: 0 / 1h: 257

缓存读取 0 33,385 34,259 35,312 33,372

输出速度 75.1 tok/s 73.3 tok/s 116.1 tok/s - 128.2 tok/s

响应

2025-02

2025-02

2025-02


2025-02

2025-02

步骤 4: identity_free text

时间 08:07:46 ~ 08:07:56 (6.9s) 08:07:25 ~ 08:07:35 (9.6s) 08:09:22 ~ 08:09:32 (6.5s) 08:09:03 ~ 08:09:13 (5.4s) 08:10:15 ~ 08:10:25 (6.4s)

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) 3,103 ms 1,732 ms 2,515 ms 2,945 ms 3,633 ms

首文本(SSE delta) 3,104 ms 1,733 ms 2,516 ms 2,946 ms 3,634 ms

总耗时 6,038 ms 8,596 ms 5,539 ms 4,109 ms 5,438 ms

输入 34,597 10 10 29 10

输出 363 987 421 45 295

缓存创建

5m: 0 / 1h: 0

5m: 0 / 1h: 99

5m: 41 / 1h: 0

111

5m: 0 / 1h: 111

105

5m: 0 / 1h: 105

缓存读取 0 33,668 34,321 35,743 33,629

输出速度 123.7 tok/s 143.8 tok/s 139.3 tok/s 38.7 tok/s 163.5 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软件工程、写作、分析等各种专业任务。

我是Claude Haiku 4.5模型，由Anthropic公司开发。作为Claude系列的轻量级版本，我为用户提供快速高效的AI助手服务，支持编程、分析和创意写作等任务。

我是Claude Haiku 4.5，由Anthropic开发的AI助手。属于Claude产品家族，用于对话、编程、分析等多种任务。


我是Claude，由Anthropic公司开发的AI助手。我可以帮助用户进行对话、编程、分析、创意写作等多种任务。

我是Claude Haiku 4.5，Anthropic公司开发的AI助手，属于Claude产品家族。

步骤 5: knowledge_recall text

时间 08:11:53 ~ 08:12:13 (19.0s) 08:09:20 ~ 08:09:40 (16.0s) 08:10:37 ~ 08:10:52 (12.9s) 08:12:43 ~ 08:12:58 (14.0s) 08:13:25 ~ 08:13:45 (16.9s)

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) 1,646 ms 1,712 ms 2,407 ms 2,077 ms 1,482 ms

首文本(SSE delta) 1,669 ms 1,714 ms 2,410 ms 2,078 ms 1,483 ms

总耗时 18,091 ms 15,002 ms 12,217 ms 12,935 ms 15,788 ms

输入 10,289 10 10 1,017 10

输出 2,486 2,181 1,731 172 1,977

缓存创建

5m: 0 / 1h: 0

9,372

5m: 0 / 1h: 9,372

8,953

5m: 0 / 1h: 8,953

8,973

5m: 0 / 1h: 8,973

2,926

5m: 0 / 1h: 2,926

缓存读取 0 0 0 0 5,721

输出速度 151.4 tok/s 164.1 tok/s 176.5 tok/s 15.8 tok/s 138.2 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


```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":"Spa

```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

总计

总输入(含缓存) 145,693 143,364 145,932 152,553 142,554

总输出 3,209 3,548 2,555 250 2,615

总缓存创建

5m: 0 / 1h: 0

43,139

5m: 0 / 1h: 43,139

77,302

5m: 34,362 / 1h: 42,940

44,882

5m: 0 / 1h: 44,882

3,538

5m: 0 / 1h: 3,538

总缓存读取 0 100,175 68,580 106,344 138,966

输出速度 139.5 tok/s 151.0 tok/s 153.0 tok/s 18.1 tok/s 136.5 tok/s

总执行时间 39.6s 41.0s 33.8s 35.0s 34.1s

总墙钟时间

11m 28.8s

含步骤间等待

8m 51.8s

含步骤间等待

10m 6.0s

含步骤间等待

12m 7.8s

含步骤间等待

12m 52.9s

含步骤间等待

按官方价目重估

同等内容若直发 Anthropic 的估算 ⓘ

$0.1617

Anthropic: Claude Haiku 4.5

$0.0817

Anthropic: Claude Haiku 4.5

$0.1002

Anthropic: Claude Haiku 4.5

$0.0693

Anthropic: Claude Haiku 4.5

$0.0314

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 one-api (100%) direct (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_01dai1EaXyjEHF7NmqLUVfB4 msg_01CgzKNtwECsZpJfLiZ66mLm msg_01BLTyRpe3EB3qNunsxQQZvU msg_01RWSHchiducMyfK3DA7i7Zv msg_01MRzLub28WY5inHE9vTup6A

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
R-ultra	cf-ray: a0c599910aade3bf-NRT server: cloudflare x-new-api-version: v0.0.0 x-oneapi-request-id: 202606160011546579222978268d9d6p6pWMDtm	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	a0c599910aade3bf-NRT
O-Max	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_011Cc5y3aQUuB23qy39kPW8k
O-Max	server: nginx x-oneapi-request-id: 202606160010385145413348268d9d6O2g09yAG	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-	cf-ray: a0c59acb59fb0e34-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	836c27fe-8ed6-45c4-a295-292923a9d6c0 a0c59acb59fb0e34-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: 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_011Cc5yMeACjQUx1GD5NKCLu

协议指纹评分（机器自动）

总分 / 100 66 99 80 88 99

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 0.0 /10 ×20 10.0 /10 ×20 5.7 /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

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 5.0 /10 ×14 10.0 /10 ×14 10.0 /10 ×14 10.0 /10 ×14 10.0 /10 ×14

sliding 5m 5.0 /10 ×13 10.0 /10 ×13 7.7 /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 5.0 /10 ×8 0.0 /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

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 10.0 /10 ×5 10.0 /10 ×5 10.0 /10 ×5

Req-ID 透传 0.0 /10 ×4 10.0 /10 ×4 0.0 /10 ×4 0.0 /10 ×4 10.0 /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 66 v3.16.0 99 v3.16.0 80 v3.16.0 88 v3.16.0 基准（参考）

雷达图例

共 21 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 中文遵循 instruction_following_lang
21 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 6964 ms 7225 ms 5798 ms 5782 ms 5765 ms

输出速度 139.5 tok/s 151.0 tok/s 153.0 tok/s 18.1 tok/s 136.5 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 ⓘ

$0.1617

$0.0817

$0.1002

$0.0693

$0.0314

缓存读取占比 ⓘ 0% 70% 47% 70% 97%

揭晓

通道

服务商 FoxCode ↗

通道R-ultra

服务商 SAIAi ↗

通道O-Max

服务商 Novaris ↗

通道O-Max

服务商 WorldBase.ai ↗

通道M-

官方基准

返回列表