模型对比 | RPDiag

测试说明点击展开 5 步 quick-probe-v1 详解

Codex/OpenAI 版快速探针：前 4 步同一会话，第 5 步（知识召回）独立新会话，覆盖服务无关的核心信号—— 1. ping/pong —— 单字指令遵循 + 建立会话。 2. 身份（结构化）—— 三行 vendor / brand / model，机器可解析（期望按官方基线多数派 baseline-derived，不硬编码 OpenAI 身份）。 3. 知识截止 —— YYYY-MM 格式，与官方基线对照。 4. 身份（自由格式）—— 自然语言自报，给包装层暴露品牌的机会。 5. 知识召回 —— 工具禁用、独立新会话的参数化公共事实题。 codex 走 /responses wire：缓存 / Anthropic 信封 / extended-thinking 等维度对其 N/A（见 codex 评分 profile），故无需 claude 版的 5 分钟 sliding-cache 边界时序，步间仅轻量间隔避免限流。

5 通道 · 5 步骤

2026/07/02 运行日期

gpt-5.4 官方基准

全部 ≥ 70 与基线相似度

已揭晓揭晓进度

步骤 / 指标

duckcoding

cx专用

right

o-cx

官方基准

iKunAPI

O-Pro20X

LyraAPI

O-Plus

基本信息

服务名 Codex Codex Codex Codex Codex

模型 gpt-5.4 gpt-5.4 gpt-5.4 gpt-5.4 gpt-5.4

客户端 codex_exec v0.139.0 codex_exec v0.139.0 codex_exec v0.139.0 codex_exec v0.139.0 codex_exec v0.139.0

偏差标红基于官方基准

步骤 1: ping text

时间 13:47:06 ~ 13:47:21 (14.5s) 13:47:08 ~ 13:47:28 (15.0s) 13:47:12 ~ 13:47:27 (12.4s) 13:47:04 ~ 13:47:19 (12.1s) 13:47:10 ~ 13:47:30 (15.6s)

提示词

查看本步 Prompt

ping. Reply with the single word: pong

HTTP 头(TTFB) 1,931 ms 822 ms 772 ms 1,392 ms 1,990 ms

总耗时 3,808 ms 4,543 ms 1,749 ms 1,791 ms 4,739 ms

输入 6,819 6,525 163 1,699 163

输出 5 17 5 17 5

缓存读取 1,920 2,432 8,576 7,040 8,576

响应

pong

pong

pong

pong

pong

步骤 2: identity text

时间 13:47:43 ~ 13:47:58 (14.8s) 13:48:09 ~ 13:48:34 (20.8s) 13:48:06 ~ 13:48:26 (16.9s) 13:47:41 ~ 13:48:01 (15.7s) 13:48:00 ~ 13:48:20 (17.8s)

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) 1,836 ms 961 ms 788 ms 2,035 ms 1,988 ms

总耗时 4,463 ms 5,020 ms 6,238 ms 5,328 ms 6,893 ms

输入 304 522 304 304 1,840

输出 113 155 227 176 217

缓存读取 8,576 8,576 8,576 8,576 7,040

输出速度 43.0 tok/s 38.2 tok/s 41.7 tok/s 53.4 tok/s 44.2 tok/s

响应

vendor: openai
brand: gpt
model: gpt-5

vendor: openai
brand: gpt
model: gpt-5

vendor: openai
brand: gpt
model: gpt-5

vendor: openai
brand: gpt
model: gpt-5

vendor: openai
brand: gpt
model: gpt-5

步骤 3: cutoff text

时间 13:48:33 ~ 13:48:48 (14.0s) 13:48:58 ~ 13:49:13 (13.3s) 13:48:46 ~ 13:49:01 (12.7s) 13:48:37 ~ 13:48:52 (12.7s) 13:48:49 ~ 13:49:09 (15.2s)

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) 2,009 ms 1,309 ms 888 ms 1,577 ms 2,455 ms

总耗时 3,208 ms 2,797 ms 1,996 ms 2,302 ms 4,476 ms

输入 348 558 4,958 340 340

输出 28 26 28 35 8

缓存读取 8,576 8,576 3,968 8,576 8,576

输出速度 - - - 48.3 tok/s -

响应

2024-06

2024-06

2024-06

2024-08

2024-06

步骤 4: identity_free text

时间 13:49:19 ~ 13:49:34 (13.4s) 13:49:43 ~ 13:49:58 (12.9s) 13:49:39 ~ 13:49:54 (13.6s) 13:49:10 ~ 13:49:25 (12.2s) 13:49:25 ~ 13:49:45 (15.4s)

提示词

查看本步 Prompt

中文做个自我介绍：你是什么产品或助手？谁开发的？50 字内。

HTTP 头(TTFB) 1,147 ms 1,211 ms 1,027 ms 1,182 ms 3,718 ms

总耗时 2,821 ms 2,528 ms 2,894 ms 1,884 ms 4,587 ms

输入 383 78 385 375 369

输出 30 32 36 35 14

缓存读取 8,576 9,088 8,576 8,576 8,576

输出速度 17.9 tok/s 24.3 tok/s 19.3 tok/s 49.9 tok/s -

响应

我是 OpenAI 开发的 ChatGPT 助手。

我是 OpenAI 打造的 GPT-5 API 助手。

我是 OpenAI 开发的 AI 助手，基于 GPT-5。

我是 OpenAI 开发的 AI 助手，基于 GPT-5。

我是 GPT-5，由 OpenAI 开发。

步骤 5: knowledge_recall text

时间 13:50:16 ~ 13:50:36 (16.4s) 13:50:21 ~ 13:50:41 (19.1s) 13:50:28 ~ 13:50:48 (15.7s) 13:49:54 ~ 13:50:15 (15.2s) 13:50:13 ~ 13:50:33 (16.5s)

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) 1,969 ms 1,490 ms 880 ms 1,488 ms 1,978 ms

总耗时 5,742 ms 8,658 ms 5,018 ms 4,830 ms 5,619 ms

输入 85 2,429 85 6,741 6,741

输出 179 338 179 179 179

缓存读取 6,656 4,608 6,656 0 0

输出速度 47.4 tok/s 47.2 tok/s 43.3 tok/s 53.6 tok/s 49.2 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

总计

总输入(含缓存) 42,243 43,392 42,247 42,227 42,221

总输出 355 568 475 442 423

总缓存创建 0 0 0 0 0

总缓存读取 34,304 33,280 36,352 32,768 32,768

输出速度 39.9 tok/s 41.9 tok/s 38.6 tok/s 52.7 tok/s 46.3 tok/s

总执行时间 1m 13.2s 1m 21.1s 1m 11.3s 1m 7.9s 1m 20.4s

总墙钟时间

3m 30.9s

含步骤间等待

3m 33.9s

含步骤间等待

3m 36.9s

含步骤间等待

3m 11.0s

含步骤间等待

3m 23.9s

含步骤间等待

按官方价目重估

同等内容若直发 Anthropic 的估算 ⓘ

$0.0337

OpenAI: GPT-5.4

$0.0421

OpenAI: GPT-5.4

$0.0310

OpenAI: GPT-5.4

$0.0385

OpenAI: GPT-5.4

$0.0382

OpenAI: GPT-5.4

通道指纹

来源步骤 knowledge_recall knowledge_recall knowledge_recall knowledge_recall knowledge_recall

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

上游 - - - - -

CDN - cloudflare - - cloudflare

ID 格式 - - - - -

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
cx专用	server: Nginx x-oneapi-request-id: 202607020550177154104588268d9d6YW5ZLn9O	—	hdr:x-oneapi-request-id	—
o-cx	via: 1.1 Caddy cf-ray: a14b5f5d4dc37963-NRT server: cloudflare	via_chain: ["1.1 Caddy"]	hdr:server:cloudflarehdr:cf-rayhdr:via	a14b5f5d4dc37963-NRT
官方基准	server: nginx/1.22.1	—	—	—
O-Pro20X	server: nginx x-new-api-version: v1.0.0-rc.12 x-oneapi-request-id: 202607020549553154891108268d9d6dHlXtzih	—	hdr:x-new-api-versionhdr:x-oneapi-request-id	—
O-Plus	cf-ray: a14b5f2eafba25c2-AMS server: cloudflare x-new-api-version: v1.0.0-rc.8 x-oneapi-request-id: 202607020550155820117978268d9d6FBjZhwGv	—	hdr:x-new-api-versionhdr:x-oneapi-request-idhdr:server:cloudflarehdr:cf-ray	a14b5f2eafba25c2-AMS

协议指纹评分（机器自动）

总分 / 100 95 89 100 98 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 10.0 /10 ×20 10.0 /10 ×20 10.0 /10 ×20 10.0 /10 ×20 10.0 /10 ×20

延迟基线 7.2 /10 ×15 10.0 /10 ×15 10.0 /10 ×15 9.0 /10 ×15 4.7 /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

系统提示纯净 10.0 /10 ×8 5.0 /10 ×8 10.0 /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

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

流式投递未计入未计入未计入未计入未计入

另有 9 个维度不适用于本服务，已排除评分（不计入总分）

综合结论

与基线相似度 95 v3.26.0 89 v3.26.0 基准（参考） 98 v3.26.0 91 v3.26.0

雷达图例

共 8 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 自由身份 identity_free_clean
4 系统提示纯净 system_prompt_clean
5 缓存命中比 cache_hit_ratio_match
6 知识召回 knowledge_recall_match
7 延迟基线 latency_baseline_match
8 结构化身份 identity_structured_match

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 4008 ms 4709 ms 3579 ms 3227 ms 5263 ms

输出速度 39.9 tok/s 41.9 tok/s 38.6 tok/s 52.7 tok/s 46.3 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 ⓘ

$0.0337

$0.0421

$0.0310

$0.0385

$0.0382

缓存读取占比 ⓘ 81% 77% 86% 78% 78%

揭晓

通道

服务商 duckcoding ↗

通道cx专用

服务商 right ↗

通道o-cx

官方基准

服务商 iKunAPI ↗

通道O-Pro20X

服务商 LyraAPI ↗

通道O-Plus

返回列表