循序渐进 · AIP 教学 · 核心平台(十三)
管理员:LLM 注册速率限制
这一篇是纯数值参照表:各注册层级在商业环境与政府环境下的 TPM(每分钟 token 数)与 RPM(每分钟请求数)上限。
本文来源 · Source
内容整理自 Palantir Foundry 官方文档:
https://www.palantir.com/docs/foundry/aip/llm-enrollment-rate-limits/
原始标题:Administration • LLM enrollment rate limits • Palantir · 所属:核心平台(AIP 总览与治理)
https://www.palantir.com/docs/foundry/aip/llm-enrollment-rate-limits/
原始标题:Administration • LLM enrollment rate limits • Palantir · 所属:核心平台(AIP 总览与治理)
★
先记住这几条
① 限制按注册层级划分
不同档位的配额不同。
② 双指标:TPM 与 RPM
token 速率和请求数速率都要满足。
③ 商业与政府环境不同
USG 环境有独立的一套数值。
0
写在前面
下表包含商业环境和美国政府(USG)环境中各注册层级的每分钟 token 数(TPM)和每分钟请求数(RPM)注册限制。对于同时启用 Azure 和 OpenAI 的注册,Azure 和 OpenAI 的注册限制将比下表所示翻倍。此外,对于地理限制在单个区域的注册,在 Large 和 X-large 层级下,TPM 和 RPM 可能低于表中所示值。
<!--start-->
如果启用了多个后端,速率限制为所有后端的总和。
| 模型名称 | 模型后端 | 按用户限制 | Small 层级 | Medium 层级 | Large 层级 | XLarge 层级 |
|---|---|---|---|---|---|---|
| Claude Haiku 4.5 | Anthropic | 1,000,000 TPM<br>100 RPM | 600,000 TPM<br>600 RPM | 1,500,000 TPM<br>750 RPM | 2,000,000 TPM<br>875 RPM | 2,500,000 TPM<br>1,000 RPM |
| Claude Haiku 4.5 | Amazon Bedrock | 1,000,000 TPM<br>100 RPM | 200,000 TPM<br>150 RPM | 1,100,000 TPM<br>300 RPM | 2,600,000 TPM<br>600 RPM | 5,100,000 TPM<br>1,100 RPM |
| Claude Haiku 4.5 | Azure | 1,000,000 TPM<br>100 RPM | 550,000 TPM<br>550 RPM | 5,000,000 TPM<br>5,000 RPM | 7,500,000 TPM<br>7,500 RPM | 10,000,000 TPM<br>10,000 RPM |
| Claude Haiku 4.5 | Google Vertex | 1,000,000 TPM<br>100 RPM | 100,000 TPM<br>50 RPM | 4,000,000 TPM<br>2,000 RPM | 6,000,000 TPM<br>3,000 RPM | 8,000,000 TPM<br>4,000 RPM |
| Claude Haiku 4.5 | Google Vertex (USG) | 1,000,000 TPM<br>100 RPM | 100,000 TPM<br>50 RPM | 4,000,000 TPM<br>2,000 RPM | 6,000,000 TPM<br>3,000 RPM | 8,000,000 TPM<br>4,000 RPM |
| Claude Opus 4.1 | Amazon Bedrock | 400,000 TPM<br>5 RPM | 100,000 TPM<br>25 RPM | 500,000 TPM<br>100 RPM | 1,000,000 TPM<br>150 RPM | 2,000,000 TPM<br>200 RPM |
| Claude Opus 4.1 | Google Vertex | 400,000 TPM<br>5 RPM | 100,000 TPM<br>25 RPM | 400,000 TPM<br>100 RPM | 600,000 TPM<br>150 RPM | 800,000 TPM<br>200 RPM |
| Claude Opus 4.5 | Anthropic | 1,000,000 TPM<br>100 RPM | 600,000 TPM<br>525 RPM | 1,500,000 TPM<br>700 RPM | 2,000,000 TPM<br>800 RPM | 2,500,000 TPM<br>900 RPM |
| Claude Opus 4.5 | Amazon Bedrock | 1,000,000 TPM<br>100 RPM | 100,000 TPM<br>25 RPM | 1,000,000 TPM<br>100 RPM | 2,000,000 TPM<br>200 RPM | 4,000,000 TPM<br>400 RPM |
| Claude Opus 4.5 | Azure | 1,000,000 TPM<br>100 RPM | 550,000 TPM<br>550 RPM | 1,500,000 TPM<br>600 RPM | 2,500,000 TPM<br>700 RPM | 3,500,000 TPM<br>800 RPM |
| Claude Opus 4.5 | Google Vertex | 1,000,000 TPM<br>100 RPM | 100,000 TPM<br>25 RPM | 1,000,000 TPM<br>100 RPM | 1,500,000 TPM<br>150 RPM | 2,000,000 TPM<br>200 RPM |
| Claude Opus 4.5 | Google Vertex (USG) | 1,000,000 TPM<br>100 RPM | 100,000 TPM<br>25 RPM | 1,000,000 TPM<br>100 RPM | 1,500,000 TPM<br>150 RPM | 2,000,000 TPM<br>200 RPM |
| Claude Opus 4.6 | Anthropic | 1,500,000 TPM<br>150 RPM | 100,000 TPM<br>10 RPM | 1,000,000 TPM<br>200 RPM | 1,500,000 TPM<br>300 RPM | 2,000,000 TPM<br>400 RPM |
| Claude Opus 4.6 | Amazon Bedrock | 1,500,000 TPM<br>150 RPM | 200,000 TPM<br>20 RPM | 3,000,000 TPM<br>300 RPM | 4,000,000 TPM<br>400 RPM | 6,000,000 TPM<br>600 RPM |
| Claude Opus 4.6 | Azure | 1,500,000 TPM<br>150 RPM | 550,000 TPM<br>550 RPM | 1,500,000 TPM<br>600 RPM | 2,500,000 TPM<br>700 RPM | 3,500,000 TPM<br>800 RPM |
| Claude Opus 4.6 | Google Vertex | 1,500,000 TPM<br>150 RPM | 200,000 TPM<br>25 RPM | 5,000,000 TPM<br>2,000 RPM | 8,000,000 TPM<br>3,000 RPM | 10,000,000 TPM<br>4,000 RPM |
| Claude Opus 4.6 | Google Vertex (USG) | 1,500,000 TPM<br>150 RPM | 200,000 TPM<br>25 RPM | 5,000,000 TPM<br>2,000 RPM | 8,000,000 TPM<br>3,000 RPM | 10,000,000 TPM<br>4,000 RPM |
| Claude Opus 4.7 | Anthropic | 1,500,000 TPM<br>150 RPM | 100,000 TPM<br>10 RPM | 1,000,000 TPM<br>200 RPM | 1,500,000 TPM<br>300 RPM | 2,000,000 TPM<br>400 RPM |
| Claude Opus 4.7 | Amazon Bedrock | 1,500,000 TPM<br>150 RPM | 200,000 TPM<br>20 RPM | 3,000,000 TPM<br>300 RPM | 4,000,000 TPM<br>400 RPM | 6,000,000 TPM<br>600 RPM |
| Claude Opus 4.7 | Azure | 1,500,000 TPM<br>150 RPM | 50,000 TPM<br>50 RPM | 1,000,000 TPM<br>100 RPM | 2,000,000 TPM<br>200 RPM | 3,000,000 TPM<br>300 RPM |
| Claude Opus 4.7 | Google Vertex | 1,500,000 TPM<br>150 RPM | 200,000 TPM<br>25 RPM | 5,000,000 TPM<br>2,000 RPM | 8,000,000 TPM<br>3,000 RPM | 10,000,000 TPM<br>4,000 RPM |
| Claude Opus 4.7 | Google Vertex (USG) | 1,500,000 TPM<br>150 RPM | 200,000 TPM<br>25 RPM | 5,000,000 TPM<br>2,000 RPM | 8,000,000 TPM<br>3,000 RPM | 10,000,000 TPM<br>4,000 RPM |
| Claude Opus 4.8 | Anthropic | 1,500,000 TPM<br>150 RPM | 100,000 TPM<br>10 RPM | 1,000,000 TPM<br>200 RPM | 1,500,000 TPM<br>300 RPM | 2,000,000 TPM<br>400 RPM |
| Claude Opus 4.8 | Amazon Bedrock | 1,500,000 TPM<br>150 RPM | 200,000 TPM<br>20 RPM | 3,000,000 TPM<br>300 RPM | 4,000,000 TPM<br>400 RPM | 6,000,000 TPM<br>600 RPM |
| Claude Opus 4.8 | Amazon Bedrock (USG) | 1,500,000 TPM<br>150 RPM | 100,000 TPM<br>25 RPM | 2,500,000 TPM<br>1,250 RPM | 5,000,000 TPM<br>2,500 RPM | 7,500,000 TPM<br>3,750 RPM |
| Claude Opus 4.8 | Azure | 1,500,000 TPM<br>150 RPM | 50,000 TPM<br>50 RPM | 1,000,000 TPM<br>100 RPM | 2,000,000 TPM<br>200 RPM | 3,000,000 TPM<br>300 RPM |
| Claude Opus 4.8 | Google Vertex | 1,500,000 TPM<br>150 RPM | 200,000 TPM<br>25 RPM | 5,000,000 TPM<br>2,000 RPM | 8,000,000 TPM<br>3,000 RPM | 10,000,000 TPM<br>4,000 RPM |
| Claude Opus 4.8 | Google Vertex (USG) | 1,500,000 TPM<br>150 RPM | 200,000 TPM<br>25 RPM | 5,000,000 TPM<br>2,000 RPM | 8,000,000 TPM<br>3,000 RPM | 10,000,000 TPM<br>4,000 RPM |
| Claude Opus 5 | Anthropic | 1,500,000 TPM<br>150 RPM | 100,000 TPM<br>10 RPM | 1,000,000 TPM<br>200 RPM | 1,500,000 TPM<br>300 RPM | 2,000,000 TPM<br>400 RPM |
| Claude Opus 5 | Amazon Bedrock | 1,500,000 TPM<br>150 RPM | 200,000 TPM<br>20 RPM | 3,000,000 TPM<br>300 RPM | 4,000,000 TPM<br>400 RPM | 6,000,000 TPM<br>600 RPM |
| Claude Opus 5 | Azure | 1,500,000 TPM<br>150 RPM | 50,000 TPM<br>50 RPM | 1,000,000 TPM<br>100 RPM | 2,000,000 TPM<br>200 RPM | 3,000,000 TPM<br>300 RPM |
| Claude Opus 5 | Google Vertex | 1,500,000 TPM<br>150 RPM | 200,000 TPM<br>25 RPM | 5,000,000 TPM<br>2,000 RPM | 8,000,000 TPM<br>3,000 RPM | 10,000,000 TPM<br>4,000 RPM |
| Claude Opus 5 | Google Vertex (USG) | 1,500,000 TPM<br>150 RPM | 200,000 TPM<br>25 RPM | 5,000,000 TPM<br>2,000 RPM | 8,000,000 TPM<br>3,000 RPM | 10,000,000 TPM<br>4,000 RPM |
| Claude Sonnet 4 | Google Vertex (USG) | 400,000 TPM<br>25 RPM | 100,000 TPM<br>25 RPM | 250,000 TPM<br>50 RPM | 375,000 TPM<br>75 RPM | 500,000 TPM<br>100 RPM |
| Claude Sonnet 4.5 | Anthropic | 1,000,000 TPM<br>100 RPM | 600,000 TPM<br>525 RPM | 1,500,000 TPM<br>700 RPM | 2,000,000 TPM<br>800 RPM | 2,500,000 TPM<br>900 RPM |
| Claude Sonnet 4.5 | Amazon Bedrock | 1,000,000 TPM<br>100 RPM | 200,000 TPM<br>125 RPM | 1,100,000 TPM<br>300 RPM | 4,100,000 TPM<br>600 RPM | 8,100,000 TPM<br>1,100 RPM |
| Claude Sonnet 4.5 | Amazon Bedrock (USG) | 1,000,000 TPM<br>100 RPM | 100,000 TPM<br>25 RPM | 5,000,000 TPM<br>2,500 RPM | 7,500,000 TPM<br>3,750 RPM | 10,000,000 TPM<br>5,000 RPM |
| Claude Sonnet 4.5 | Azure | 1,000,000 TPM<br>100 RPM | 550,000 TPM<br>550 RPM | 1,500,000 TPM<br>600 RPM | 2,500,000 TPM<br>700 RPM | 3,500,000 TPM<br>800 RPM |
| Claude Sonnet 4.5 | Google Vertex | 1,000,000 TPM<br>100 RPM | 100,000 TPM<br>25 RPM | 1,000,000 TPM<br>500 RPM | 1,500,000 TPM<br>750 RPM | 2,000,000 TPM<br>1,000 RPM |
| Claude Sonnet 4.5 | Google Vertex (USG) | 1,000,000 TPM<br>100 RPM | 100,000 TPM<br>25 RPM | 500,000 TPM<br>500 RPM | 750,000 TPM<br>750 RPM | 1,000,000 TPM<br>1,000 RPM |
| Claude Sonnet 4.6 | Anthropic | 1,000,000 TPM<br>100 RPM | 600,000 TPM<br>510 RPM | 1,500,000 TPM<br>700 RPM | 2,000,000 TPM<br>800 RPM | 2,500,000 TPM<br>900 RPM |
| Claude Sonnet 4.6 | Amazon Bedrock | 1,000,000 TPM<br>100 RPM | 300,000 TPM<br>120 RPM | 3,000,000 TPM<br>750 RPM | 6,000,000 TPM<br>1,500 RPM | 9,000,000 TPM<br>2,250 RPM |
| Claude Sonnet 4.6 | Azure | 1,000,000 TPM<br>100 RPM | 550,000 TPM<br>550 RPM | 1,500,000 TPM<br>600 RPM | 2,500,000 TPM<br>700 RPM | 3,500,000 TPM<br>800 RPM |
| Claude Sonnet 4.6 | Google Vertex | 1,000,000 TPM<br>100 RPM | 200,000 TPM<br>20 RPM | 4,000,000 TPM<br>2,000 RPM | 6,000,000 TPM<br>3,000 RPM | 8,000,000 TPM<br>4,000 RPM |
| Claude Sonnet 4.6 | Google Vertex (USG) | 1,000,000 TPM<br>100 RPM | 100,000 TPM<br>25 RPM | 4,000,000 TPM<br>2,000 RPM | 6,000,000 TPM<br>3,000 RPM | 8,000,000 TPM<br>4,000 RPM |
| Claude Sonnet 5 | Anthropic | 1,000,000 TPM<br>100 RPM | 600,000 TPM<br>510 RPM | 1,500,000 TPM<br>700 RPM | 2,000,000 TPM<br>800 RPM | 2,500,000 TPM<br>900 RPM |
| Claude Sonnet 5 | Amazon Bedrock | 1,000,000 TPM<br>100 RPM | 300,000 TPM<br>120 RPM | 3,000,000 TPM<br>750 RPM | 6,000,000 TPM<br>1,500 RPM | 9,000,000 TPM<br>2,250 RPM |
| Claude Sonnet 5 | Amazon Bedrock (USG) | 1,000,000 TPM<br>100 RPM | 100,000 TPM<br>25 RPM | 1,000,000 TPM<br>500 RPM | 2,000,000 TPM<br>1,000 RPM | 3,000,000 TPM<br>1,500 RPM |
| Claude Sonnet 5 | Azure | 1,000,000 TPM<br>100 RPM | 1,050,000 TPM<br>1,050 RPM | 2,500,000 TPM<br>1,250 RPM | 5,000,000 TPM<br>2,500 RPM | 7,500,000 TPM<br>3,750 RPM |
| Claude Sonnet 5 | Google Vertex | 1,000,000 TPM<br>100 RPM | 200,000 TPM<br>25 RPM | 4,000,000 TPM<br>2,000 RPM | 6,000,000 TPM<br>3,000 RPM | 8,000,000 TPM<br>4,000 RPM |
| Claude Sonnet 5 | Google Vertex (USG) | 1,000,000 TPM<br>100 RPM | 200,000 TPM<br>25 RPM | 4,000,000 TPM<br>2,000 RPM | 6,000,000 TPM<br>3,000 RPM | 8,000,000 TPM<br>4,000 RPM |
| Gemini 3 Flash (Preview) | Google Vertex | 2,000,000 TPM<br>400 RPM | 600,000 TPM<br>600 RPM | 6,500,000 TPM<br>1,400 RPM | 9,500,000 TPM<br>2,300 RPM | 12,500,000 TPM<br>3,500 RPM |
| Gemini 3 Flash (Preview) | Google Vertex (USG) | 2,000,000 TPM<br>400 RPM | 100,000 TPM<br>25 RPM | 1,000,000 TPM<br>300 RPM | 1,500,000 TPM<br>600 RPM | 2,000,000 TPM<br>1,000 RPM |
| Gemini 3.1 Flash-Lite | Google Vertex | 2,000,000 TPM<br>400 RPM | 600,000 TPM<br>600 RPM | 6,500,000 TPM<br>1,400 RPM | 9,500,000 TPM<br>2,300 RPM | 12,500,000 TPM<br>3,500 RPM |
| Gemini 3.1 Flash-Lite | Google Vertex (USG) | 2,000,000 TPM<br>400 RPM | 100,000 TPM<br>100 RPM | 6,000,000 TPM<br>900 RPM | 9,000,000 TPM<br>1,800 RPM | 12,000,000 TPM<br>3,000 RPM |
| Gemini 3.1 Pro (Preview) | Google Vertex | 2,000,000 TPM<br>400 RPM | 1,000,000 TPM<br>600 RPM | 7,000,000 TPM<br>1,500 RPM | 10,000,000 TPM<br>2,250 RPM | 15,000,000 TPM<br>3,750 RPM |
| Gemini 3.1 Pro (Preview) | Google Vertex (USG) | 2,000,000 TPM<br>400 RPM | 1,000,000 TPM<br>600 RPM | 7,000,000 TPM<br>1,500 RPM | 10,000,000 TPM<br>2,250 RPM | 15,000,000 TPM<br>3,750 RPM |
| Gemini 3.5 Flash | Google Vertex | 2,000,000 TPM<br>400 RPM | 1,000,000 TPM<br>600 RPM | 6,500,000 TPM<br>1,400 RPM | 9,500,000 TPM<br>2,300 RPM | 12,500,000 TPM<br>3,500 RPM |
| Gemini 3.5 Flash | Google Vertex (USG) | 2,000,000 TPM<br>400 RPM | 500,000 TPM<br>100 RPM | 6,000,000 TPM<br>900 RPM | 9,000,000 TPM<br>1,800 RPM | 12,000,000 TPM<br>3,000 RPM |
| Gemini 3.5 Flash-Lite | Google Vertex | 2,000,000 TPM<br>400 RPM | 1,000,000 TPM<br>600 RPM | 7,000,000 TPM<br>3,000 RPM | 10,000,000 TPM<br>4,500 RPM | 15,000,000 TPM<br>7,500 RPM |
| Gemini 3.5 Flash-Lite | Google Vertex (USG) | 2,000,000 TPM<br>400 RPM | 1,000,000 TPM<br>600 RPM | 7,000,000 TPM<br>3,000 RPM | 10,000,000 TPM<br>4,500 RPM | 15,000,000 TPM<br>7,500 RPM |
| Gemini 3.6 Flash | Google Vertex | 2,000,000 TPM<br>400 RPM | 1,000,000 TPM<br>600 RPM | 7,000,000 TPM<br>3,000 RPM | 10,000,000 TPM<br>4,500 RPM | 15,000,000 TPM<br>7,500 RPM |
| Gemini 3.6 Flash | Google Vertex (USG) | 2,000,000 TPM<br>400 RPM | 1,000,000 TPM<br>600 RPM | 7,000,000 TPM<br>3,000 RPM | 10,000,000 TPM<br>4,500 RPM | 15,000,000 TPM<br>7,500 RPM |
| Gemini 3.7 Flash | Google Vertex | 2,000,000 TPM<br>400 RPM | 1,000,000 TPM<br>600 RPM | 7,000,000 TPM<br>3,000 RPM | 10,000,000 TPM<br>4,500 RPM | 15,000,000 TPM<br>7,500 RPM |
| Gemini 3.7 Flash | Google Vertex (USG) | 2,000,000 TPM<br>400 RPM | 1,000,000 TPM<br>600 RPM | 7,000,000 TPM<br>3,000 RPM | 10,000,000 TPM<br>4,500 RPM | 15,000,000 TPM<br>7,500 RPM |
| Llama 3.1 8b Instruct | Amazon Bedrock | 50,000 TPM<br>100 RPM | 100,000 TPM<br>100 RPM | 300,000 TPM<br>450 RPM | 450,000 TPM<br>675 RPM | 600,000 TPM<br>900 RPM |
| Llama 3.1 8b Instruct | Palantir-hosted | 50,000 TPM<br>100 RPM | 100,000 TPM<br>100 RPM | 300,000 TPM<br>450 RPM | 450,000 TPM<br>675 RPM | 600,000 TPM<br>900 RPM |
| Llama 3.3 70b Instruct | Amazon Bedrock | 50,000 TPM<br>100 RPM | 100,000 TPM<br>25 RPM | 300,000 TPM<br>450 RPM | 450,000 TPM<br>675 RPM | 600,000 TPM<br>900 RPM |
| Llama 3.3 70b Instruct | Palantir-hosted | 50,000 TPM<br>100 RPM | 100,000 TPM<br>25 RPM | 300,000 TPM<br>450 RPM | 450,000 TPM<br>675 RPM | 600,000 TPM<br>900 RPM |
| Llama 4 Maverick 17b 128E Instruct | Amazon Bedrock | 100,000 TPM<br>100 RPM | 100,000 TPM<br>25 RPM | 1,000,000 TPM<br>500 RPM | 2,000,000 TPM<br>1,000 RPM | 4,000,000 TPM<br>2,000 RPM |
| Llama 4 Scout 17b 16E Instruct | Amazon Bedrock | 100,000 TPM<br>100 RPM | 100,000 TPM<br>100 RPM | 1,000,000 TPM<br>500 RPM | 2,000,000 TPM<br>1,000 RPM | 4,000,000 TPM<br>2,000 RPM |
| Llama 4 Scout 17b 16E Instruct | Palantir-hosted | 100,000 TPM<br>100 RPM | 100,000 TPM<br>100 RPM | 300,000 TPM<br>450 RPM | 450,000 TPM<br>675 RPM | 600,000 TPM<br>900 RPM |
| Llama 3.2 NV EmbedQA 1B v2 | Palantir-hosted | 50,000 TPM<br>100 RPM | 60,000 TPM<br>150 RPM | 300,000 TPM<br>450 RPM | 450,000 TPM<br>675 RPM | 600,000 TPM<br>900 RPM |
| Llama 3.3 Nemotron Super 49b v1.5 | Palantir-hosted | 50,000 TPM<br>100 RPM | 100,000 TPM<br>25 RPM | 300,000 TPM<br>450 RPM | 450,000 TPM<br>675 RPM | 600,000 TPM<br>900 RPM |
| NVIDIA Nemotron 3 Nano 30B | Amazon Bedrock | 50,000 TPM<br>100 RPM | 100,000 TPM<br>25 RPM | 500,000 TPM<br>100 RPM | 1,000,000 TPM<br>150 RPM | 2,000,000 TPM<br>200 RPM |
| NVIDIA Nemotron 3 Super 120B | Amazon Bedrock | 500,000 TPM<br>100 RPM | 40,000 TPM<br>10 RPM | 1,000,000 TPM<br>200 RPM | 2,000,000 TPM<br>300 RPM | 4,000,000 TPM<br>400 RPM |
| GPT-4.1 | Azure | 400,000 TPM<br>1,000 RPM | 600,000 TPM<br>525 RPM | 2,000,000 TPM<br>1,500 RPM | 3,500,000 TPM<br>2,500 RPM | 5,500,000 TPM<br>4,500 RPM |
| GPT-4.1 | Azure (USG) | 400,000 TPM<br>1,000 RPM | 100,000 TPM<br>25 RPM | 2,000,000 TPM<br>2,000 RPM | 4,000,000 TPM<br>4,000 RPM | 6,000,000 TPM<br>6,000 RPM |
| GPT-4.1 | OpenAI | 400,000 TPM<br>1,000 RPM | 600,000 TPM<br>525 RPM | 2,000,000 TPM<br>1,500 RPM | 3,500,000 TPM<br>2,500 RPM | 5,500,000 TPM<br>4,500 RPM |
| GPT-4.1 mini | Azure | 1,000,000 TPM<br>1,000 RPM | 5,100,000 TPM<br>5,100 RPM | 15,000,000 TPM<br>7,500 RPM | 35,000,000 TPM<br>12,500 RPM | 55,000,000 TPM<br>17,500 RPM |
| GPT-4.1 mini | Azure (USG) | 1,000,000 TPM<br>1,000 RPM | 100,000 TPM<br>100 RPM | 2,000,000 TPM<br>2,000 RPM | 3,000,000 TPM<br>3,000 RPM | 4,000,000 TPM<br>4,000 RPM |
| GPT-4.1 mini | OpenAI | 1,000,000 TPM<br>1,000 RPM | 1,100,000 TPM<br>1,100 RPM | 3,000,000 TPM<br>2,000 RPM | 4,000,000 TPM<br>3,000 RPM | 6,000,000 TPM<br>5,000 RPM |
| GPT-4.1 nano | Azure | 1,000,000 TPM<br>1,000 RPM | 5,100,000 TPM<br>5,100 RPM | 15,000,000 TPM<br>7,500 RPM | 35,000,000 TPM<br>12,500 RPM | 55,000,000 TPM<br>17,500 RPM |
| GPT-4o | Azure | 400,000 TPM<br>800 RPM | 100,000 TPM<br>25 RPM | 1,000,000 TPM<br>1,000 RPM | 1,500,000 TPM<br>2,000 RPM | 3,000,000 TPM<br>4,000 RPM |
| GPT-4o | Azure (USG) | 400,000 TPM<br>800 RPM | 100,000 TPM<br>25 RPM | 1,000,000 TPM<br>1,000 RPM | 1,500,000 TPM<br>2,000 RPM | 3,000,000 TPM<br>4,000 RPM |
| GPT-4o | OpenAI | 400,000 TPM<br>800 RPM | 100,000 TPM<br>25 RPM | 1,000,000 TPM<br>1,000 RPM | 1,500,000 TPM<br>2,000 RPM | 3,000,000 TPM<br>4,000 RPM |
| GPT-5 | Azure | 1,000,000 TPM<br>1,000 RPM | 600,000 TPM<br>525 RPM | 3,500,000 TPM<br>1,500 RPM | 5,500,000 TPM<br>3,000 RPM | 10,500,000 TPM<br>5,500 RPM |
| GPT-5 Codex | Azure | 1,000,000 TPM<br>1,000 RPM | 100,000 TPM<br>25 RPM | 2,000,000 TPM<br>1,000 RPM | 3,000,000 TPM<br>2,000 RPM | 5,000,000 TPM<br>4,000 RPM |
| GPT-5 mini | Azure | 1,000,000 TPM<br>1,000 RPM | 600,000 TPM<br>600 RPM | 10,500,000 TPM<br>5,500 RPM | 20,500,000 TPM<br>10,500 RPM | 30,500,000 TPM<br>15,500 RPM |
| GPT-5 nano | Azure | 1,000,000 TPM<br>1,000 RPM | 1,100,000 TPM<br>1,100 RPM | 11,000,000 TPM<br>6,000 RPM | 31,000,000 TPM<br>16,000 RPM | 51,000,000 TPM<br>26,000 RPM |
| GPT-5.1 | Azure | 1,000,000 TPM<br>2,000 RPM | 100,000 TPM<br>25 RPM | 2,000,000 TPM<br>500 RPM | 4,000,000 TPM<br>1,000 RPM | 6,000,000 TPM<br>2,000 RPM |
| GPT-5.1 | Azure (USG) | 1,000,000 TPM<br>2,000 RPM | 100,000 TPM<br>25 RPM | 3,000,000 TPM<br>3,000 RPM | 5,250,000 TPM<br>5,250 RPM | 7,500,000 TPM<br>7,500 RPM |
| GPT-5.1 | OpenAI | 1,000,000 TPM<br>2,000 RPM | 100,000 TPM<br>25 RPM | 1,500,000 TPM<br>1,000 RPM | 3,000,000 TPM<br>2,000 RPM | 5,000,000 TPM<br>4,000 RPM |
| GPT-5.1 Codex | Azure | 1,000,000 TPM<br>1,000 RPM | 100,000 TPM<br>25 RPM | 2,000,000 TPM<br>1,000 RPM | 3,000,000 TPM<br>2,000 RPM | 4,000,000 TPM<br>4,000 RPM |
| GPT-5.1 Codex mini | Azure | 1,000,000 TPM<br>500 RPM | 100,000 TPM<br>100 RPM | 2,000,000 TPM<br>1,000 RPM | 3,000,000 TPM<br>2,000 RPM | 5,000,000 TPM<br>4,000 RPM |
| GPT-5.2 | Azure | 500,000 TPM<br>1,000 RPM | 250,000 TPM<br>50 RPM | 2,000,000 TPM<br>500 RPM | 4,000,000 TPM<br>1,000 RPM | 6,000,000 TPM<br>2,000 RPM |
| GPT-5.2 | OpenAI | 500,000 TPM<br>1,000 RPM | 250,000 TPM<br>50 RPM | 3,000,000 TPM<br>1,500 RPM | 6,000,000 TPM<br>3,000 RPM | 10,000,000 TPM<br>5,000 RPM |
| GPT-5.2 Pro | OpenAI | 400,000 TPM<br>1,000 RPM | 100,000 TPM<br>25 RPM | 1,500,000 TPM<br>1,000 RPM | 3,000,000 TPM<br>2,000 RPM | 5,000,000 TPM<br>4,000 RPM |
| GPT-5.3 Codex | Azure | 1,000,000 TPM<br>1,000 RPM | 100,000 TPM<br>100 RPM | 4,000,000 TPM<br>2,000 RPM | 6,000,000 TPM<br>4,000 RPM | 8,000,000 TPM<br>8,000 RPM |
| GPT-5.3 Codex | OpenAI | 1,000,000 TPM<br>1,000 RPM | 100,000 TPM<br>25 RPM | 3,000,000 TPM<br>1,000 RPM | 4,000,000 TPM<br>2,000 RPM | 5,000,000 TPM<br>4,000 RPM |
| GPT-5.4 | Azure | 1,000,000 TPM<br>1,000 RPM | 250,000 TPM<br>50 RPM | 4,000,000 TPM<br>2,000 RPM | 6,000,000 TPM<br>3,000 RPM | 8,000,000 TPM<br>4,000 RPM |
| GPT-5.4 | OpenAI | 1,000,000 TPM<br>1,000 RPM | 250,000 TPM<br>50 RPM | 3,000,000 TPM<br>1,500 RPM | 6,000,000 TPM<br>3,000 RPM | 10,000,000 TPM<br>5,000 RPM |
| GPT-5.4 Pro | OpenAI | 400,000 TPM<br>1,000 RPM | 100,000 TPM<br>25 RPM | 1,500,000 TPM<br>1,000 RPM | 3,000,000 TPM<br>2,000 RPM | 5,000,000 TPM<br>4,000 RPM |
| GPT-5.4 mini | Azure | 1,000,000 TPM<br>1,000 RPM | 100,000 TPM<br>100 RPM | 4,500,000 TPM<br>2,250 RPM | 9,000,000 TPM<br>4,500 RPM | 15,000,000 TPM<br>7,500 RPM |
| GPT-5.4 mini | OpenAI | 1,000,000 TPM<br>1,000 RPM | 100,000 TPM<br>100 RPM | 3,000,000 TPM<br>1,500 RPM | 6,000,000 TPM<br>3,000 RPM | 10,000,000 TPM<br>5,000 RPM |
| GPT-5.4 nano | Azure | 1,000,000 TPM<br>1,000 RPM | 100,000 TPM<br>100 RPM | 4,500,000 TPM<br>2,250 RPM | 9,000,000 TPM<br>4,500 RPM | 15,000,000 TPM<br>7,500 RPM |
| GPT-5.4 nano | OpenAI | 1,000,000 TPM<br>1,000 RPM | 100,000 TPM<br>100 RPM | 3,000,000 TPM<br>1,500 RPM | 6,000,000 TPM<br>3,000 RPM | 10,000,000 TPM<br>5,000 RPM |
| GPT-5.5 | Azure | 1,000,000 TPM<br>1,000 RPM | 250,000 TPM<br>50 RPM | 4,000,000 TPM<br>2,000 RPM | 6,000,000 TPM<br>3,000 RPM | 8,000,000 TPM<br>4,000 RPM |
| GPT-5.5 | OpenAI | 1,000,000 TPM<br>1,000 RPM | 250,000 TPM<br>50 RPM | 3,000,000 TPM<br>1,500 RPM | 6,000,000 TPM<br>3,000 RPM | 10,000,000 TPM<br>5,000 RPM |
| GPT-5.6 Luna | Azure | 3,000,000 TPM<br>1,500 RPM | 250,000 TPM<br>125 RPM | 5,000,000 TPM<br>2,500 RPM | 6,000,000 TPM<br>3,000 RPM | 7,000,000 TPM<br>3,500 RPM |
| GPT-5.6 Luna | OpenAI | 3,000,000 TPM<br>1,500 RPM | 250,000 TPM<br>50 RPM | 5,000,000 TPM<br>2,500 RPM | 10,000,000 TPM<br>5,000 RPM | 15,000,000 TPM<br>7,500 RPM |
| GPT-5.6 Sol | Azure | 5,000,000 TPM<br>2,500 RPM | 250,000 TPM<br>125 RPM | 9,000,000 TPM<br>4,500 RPM | 10,000,000 TPM<br>5,000 RPM | 11,000,000 TPM<br>5,500 RPM |
| GPT-5.6 Sol | OpenAI | 5,000,000 TPM<br>2,500 RPM | 250,000 TPM<br>50 RPM | 10,000,000 TPM<br>5,000 RPM | 20,000,000 TPM<br>10,000 RPM | 30,000,000 TPM<br>15,000 RPM |
| GPT-5.6 Terra | Azure | 3,000,000 TPM<br>1,500 RPM | 250,000 TPM<br>125 RPM | 5,000,000 TPM<br>2,500 RPM | 6,000,000 TPM<br>3,000 RPM | 7,000,000 TPM<br>3,500 RPM |
| GPT-5.6 Terra | OpenAI | 3,000,000 TPM<br>1,500 RPM | 250,000 TPM<br>50 RPM | 5,000,000 TPM<br>2,500 RPM | 10,000,000 TPM<br>5,000 RPM | 15,000,000 TPM<br>7,500 RPM |
| GPT-OSS-120B | Palantir-hosted | 50,000 TPM<br>100 RPM | 100,000 TPM<br>25 RPM | 300,000 TPM<br>450 RPM | 450,000 TPM<br>675 RPM | 600,000 TPM<br>900 RPM |
| GPT-OSS-20B | Palantir-hosted | 50,000 TPM<br>100 RPM | 100,000 TPM<br>100 RPM | 300,000 TPM<br>450 RPM | 450,000 TPM<br>675 RPM | 600,000 TPM<br>900 RPM |
| Text Embedding 3 Large | Azure | 1,000,000 TPM<br>1,500 RPM | 1,060,000 TPM<br>1,400 RPM | 2,250,000 TPM<br>3,500 RPM | 3,500,000 TPM<br>4,750 RPM | 4,750,000 TPM<br>8,500 RPM |
| Text Embedding 3 Large | Azure (USG) | 1,000,000 TPM<br>1,500 RPM | 60,000 TPM<br>400 RPM | 1,000,000 TPM<br>2,000 RPM | 2,000,000 TPM<br>3,000 RPM | 3,000,000 TPM<br>6,000 RPM |
| Text Embedding 3 Large | OpenAI | 1,000,000 TPM<br>1,500 RPM | 60,000 TPM<br>400 RPM | 1,250,000 TPM<br>2,500 RPM | 2,500,000 TPM<br>3,750 RPM | 3,750,000 TPM<br>7,500 RPM |
| Text Embedding 3 Small | Azure | 1,000,000 TPM<br>1,500 RPM | 1,060,000 TPM<br>1,400 RPM | 1,625,000 TPM<br>3,500 RPM | 2,250,000 TPM<br>4,750 RPM | 2,875,000 TPM<br>8,500 RPM |
| Text Embedding 3 Small | Azure (USG) | 1,000,000 TPM<br>1,500 RPM | 60,000 TPM<br>400 RPM | 500,000 TPM<br>2,000 RPM | 1,000,000 TPM<br>3,000 RPM | 1,500,000 TPM<br>6,000 RPM |
| Text Embedding 3 Small | OpenAI | 1,000,000 TPM<br>1,500 RPM | 60,000 TPM<br>400 RPM | 625,000 TPM<br>2,500 RPM | 1,250,000 TPM<br>3,750 RPM | 1,875,000 TPM<br>7,500 RPM |
| Whisper 1 | OpenAI | 1 TPM<br>50 RPM | 1 TPM<br>100 RPM | 2 TPM<br>500 RPM | 3 TPM<br>1,000 RPM | 4 TPM<br>2,000 RPM |
| o1 | Azure | 600,000 TPM<br>5 RPM | 100,000 TPM<br>25 RPM | 250,000 TPM<br>50 RPM | 400,000 TPM<br>60 RPM | 750,000 TPM<br>75 RPM |
| o3 | Azure | 400,000 TPM<br>100 RPM | 600,000 TPM<br>525 RPM | 1,500,000 TPM<br>1,500 RPM | 2,500,000 TPM<br>2,500 RPM | 4,500,000 TPM<br>4,500 RPM |
| o3-mini | Azure (USG) | 300,000 TPM<br>30 RPM | 100,000 TPM<br>25 RPM | 1,000,000 TPM<br>100 RPM | 2,000,000 TPM<br>200 RPM | 4,000,000 TPM<br>400 RPM |
| o4-mini | Azure | 300,000 TPM<br>100 RPM | 1,100,000 TPM<br>1,025 RPM | 2,000,000 TPM<br>2,000 RPM | 3,000,000 TPM<br>3,000 RPM | 5,000,000 TPM<br>5,000 RPM |
| text-embedding-ada-002 | Azure | 1,000,000 TPM<br>1,500 RPM | 1,450,000 TPM<br>1,450 RPM | 3,625,000 TPM<br>3,625 RPM | 4,875,000 TPM<br>4,875 RPM | 6,250,000 TPM<br>6,250 RPM |
| text-embedding-ada-002 | Azure (USG) | 1,000,000 TPM<br>1,500 RPM | 450,000 TPM<br>450 RPM | 3,000,000 TPM<br>3,500 RPM | 4,000,000 TPM<br>4,500 RPM | 5,000,000 TPM<br>5,500 RPM |
| text-embedding-ada-002 | OpenAI | 1,000,000 TPM<br>1,500 RPM | 950,000 TPM<br>950 RPM | 3,125,000 TPM<br>3,125 RPM | 4,437,500 TPM<br>4,438 RPM | 5,750,000 TPM<br>5,750 RPM |
| Document Information Extraction | Palantir-hosted | 1,000,000 TPM<br>40 RPM | 1,000,000 TPM<br>40 RPM | 1,500,000 TPM<br>300 RPM | 2,000,000 TPM<br>450 RPM | 3,000,000 TPM<br>600 RPM |
| Gemma 4 26B A4B | Amazon Bedrock | 50,000 TPM<br>100 RPM | 250,000 TPM<br>25 RPM | 5,000,000 TPM<br>500 RPM | 10,000,000 TPM<br>1,000 RPM | 20,000,000 TPM<br>2,000 RPM |
| Gemma 4 26B A4B | Palantir-hosted | 50,000 TPM<br>100 RPM | 100,000 TPM<br>25 RPM | 300,000 TPM<br>450 RPM | 450,000 TPM<br>675 RPM | 600,000 TPM<br>900 RPM |
| Schematic 7B | Palantir-hosted | 50,000 TPM<br>100 RPM | 60,000 TPM<br>150 RPM | 300,000 TPM<br>450 RPM | 450,000 TPM<br>675 RPM | 600,000 TPM<br>900 RPM |
| Whisper Large V3 | Palantir-hosted | 50,000 TPM<br>100 RPM | 60,000 TPM<br>150 RPM | 300,000 TPM<br>450 RPM | 450,000 TPM<br>675 RPM | 600,000 TPM<br>900 RPM |
| Snowflake Arctic Embed Medium | Palantir-hosted | 500,000 TPM<br>500 RPM | 60,000 TPM<br>150 RPM | 300,000 TPM<br>450 RPM | 450,000 TPM<br>675 RPM | 600,000 TPM<br>900 RPM |
| Grok 4.3 | xAI | 500,000 TPM<br>100 RPM | 50,000 TPM<br>20 RPM | 1,000,000 TPM<br>200 RPM | 1,500,000 TPM<br>300 RPM | 2,000,000 TPM<br>400 RPM |
| Grok 4.5 | xAI | 1,000,000 TPM<br>200 RPM | 100,000 TPM<br>40 RPM | 3,000,000 TPM<br>800 RPM | 4,000,000 TPM<br>1,200 RPM | 5,000,000 TPM<br>1,600 RPM |
| Grok 420 Non-Reasoning Latest | xAI | 500,000 TPM<br>100 RPM | 50,000 TPM<br>20 RPM | 1,000,000 TPM<br>200 RPM | 1,500,000 TPM<br>300 RPM | 2,000,000 TPM<br>400 RPM |
| Grok 420 Reasoning Latest | xAI | 500,000 TPM<br>100 RPM | 50,000 TPM<br>20 RPM | 1,000,000 TPM<br>200 RPM | 1,500,000 TPM<br>300 RPM | 2,000,000 TPM<br>400 RPM |
| Grok Build 0.1 | xAI | 400,000 TPM<br>100 RPM | 50,000 TPM<br>20 RPM | 1,000,000 TPM<br>200 RPM | 1,500,000 TPM<br>300 RPM | 2,000,000 TPM<br>400 RPM |
<!--end-->
延伸阅读 · 相关页面
按主题横向跳转,不必顺着目录一篇篇读。
本组其他页面 · 核心平台(AIP 总览与治理)
同一主题下的相邻内容。
- AIP 总览:把 AI 接进你的数据与运营AIP(Artificial Intelligence Platform)不是孤立的模型平台,它长在你的数据与 Onto
- AIP 能力清单:平台各处的 AI 特性平台里几乎每个应用都配备了 AIP 驱动能力。这一篇把它们按应用分类列全,作为你的能力索引表。
- 开始使用 AIP:学习路径这一篇很短,只回答一个问题:该按什么顺序学 AIP。
- 提示词工程最佳实践写提示词(prompt)这件事,直接决定 LLM 输出的质量。这一篇讲怎么写出稳定可靠的提示词,是所有 AI 应用的基本
- 平台支持的 LLM 清单AIP 支持来自 xAI、OpenAI、Anthropic、Meta、Google 等提供商的多种 LLM 与文本嵌入模
- 兼容各提供商的原生 API 端点Foundry 为主流 LLM 提供商提供代理端点,按各提供商原生 API 的格式接收请求 —— 你可以继续用熟悉的开源
- AI 伦理与治理Palantir 把负责任 AI 当作构建方式本身,而不是事后补丁。这一篇讲他们的伦理原则与治理机制。
- AIP 的安全与隐私把 AI 接进企业数据,安全是第一道门槛。这一篇讲 AIP 如何保护客户数据的隐私与安全。
- AIP 的计算用量与计费LLM 按 token 计费:输入文本和输出文本都要算钱。这一篇讲清用量从哪来、怎么算、怎么控。
- Ontology 与 AIP 的可观测性AI 跑出问题了,怎么查?这一篇讲 Workflow Lineage 里的一组能力,让你看清每条 AI 流程的来龙去脉。
- 管理员:开启 AIP 能力AIP 能力默认可能并未全部开启。这一篇是管理员的操作指南:在哪些界面、按什么顺序把能力开出来。
- 管理员:LLM 容量管理LLM 容量在行业层面是有限资源,所有提供商都会限制账户的最大可用容量。这一篇讲 AIP 如何在组织内分配这份稀缺资源。