Compute usage with AIP（AIP 计算用量）¶
AIP compute usage involves large language models (LLMs). Fundamentally, LLMs take text as an input and respond with text as an output. The amount of text input and output is measured in tokens. Compute usage for LLMs is measured in compute-seconds per some number of tokens. Different models may have different rates for compute usage, as described below.
Tokens in AIP¶
Tokens are the basic units of text that LLMs use to process and understand input. A token can be as short as a single character or as long as a whole word depending on the language and the specific model.
Importantly, tokens do not map one-to-one with words. For example, common words might be a single token, but longer or less common words may be split into multiple tokens. Even punctuation marks and spaces can be considered tokens.
Different model providers have distinct definitions for what constitutes a token; for instance, OpenAI ↗ and Anthropic ↗. On average, tokens are around 4 characters long, with a character being a single letter or punctuation mark.
In AIP, tokens are consumed by applications that send prompts to and receive prompts from LLMs. Each of these prompts and responses consist of a measurable number of tokens. These tokens can be sent to multiple LLM providers; due to differences between providers, these tokens are converted into compute-seconds to match the price of the underlying model provider.
All applications that provide LLM-backed capabilities consume tokens when being used. See the following list for the set of applications that may use tokens when you interact with their LLM-backed capabilities.
- AIP Assist
- AIP Logic
- AIP Error Enhancer
- AIP Code Assist
- AIP Analyst
- AI FDE
- Workshop LLM-backed tools
- Quiver LLM-backed tools
- Pipeline Builder LLM-backed tools
- Direct calls to the Language Model Service (including both Python and TypeScript libraries)
AIP routes text directly to backing LLMs which run the tokenization themselves. The size of the text will dictate the amount of compute that is used by the backing model to serve the response.
Take the following example sentence that is sent to the GPT-4o model.
AIP incorporates all of Palantir's advanced security measures for the protection of sensitive data in compliance with industry regulations.
This sentence contains 140 characters and will tokenize in the following way, with a | character separating each token. Note that a token is not always equivalent to a word; some words are broken into multiple tokens, like AIP and Palantir in the example below.
A|IP| incorporates| all| of| Pal|ant|ir|'s| advanced| security| measures| for| the| protection| of| sensitive| data| in| compliance| with| industry| regulations|.
This sentence contains 24 tokens and will use the following number of compute-seconds:
compute-seconds = 24 tokens * 43 compute-seconds / 10,000 tokens
compute-seconds = 24 * 43 / 10,000
compute-seconds = 0.1032
The number of tokens and characters in the above sentence was verified with OpenAI's Tokenizer feature ↗.
Understanding drivers of compute usage with AIP¶
Usage of compute-seconds resulting from LLM tokens is attached directly to the individual application resource that requests the usage. For example, if you use AIP to automatically explain a pipeline in Pipeline Builder, the compute-seconds used by the LLM to generate that explanation will be attributed to that specific pipeline. This is true across the platform; keeping this in mind will help you track where you are using tokens.
In some cases, compute usage is not attributable to a single resource in the platform; examples include AIP Assist and Error Explainer, among others. When usage is not attributable to a single resource, the tokens will be attributed to the user folder initiating the use of tokens.
We recommend staying aware of the tokens that are sent to LLMs on your behalf. Generally, the more information that you include when using LLMs, the more compute-seconds will be used. For example, the following scenarios describe different ways of using compute-seconds.
- In Pipeline Builder, you can ask AIP to explain your transformation nodes; the number of selected nodes affects the number of tokens used by the LLM to generate a response, and thus compute-second usage. This is because as the number of nodes increases, so does the amount of text the LLM must process regarding the configuration of those nodes.
- In AIP Assist, asking the LLM to generate large blocks of code requires more output tokens. Shorter responses use fewer tokens and thus less compute.
- In AIP Logic, sending large amounts of text with your prompts requires more tokens and thus more compute-seconds.
Exporting AIP token usage data¶
To analyze your enrollment's LLM usage in detail, you can export the AIP Token Usage dataset from the Internal dataset export section in Control Panel. This dataset provides daily breakdowns of token consumption by model and resource, along with the corresponding usage in compute-seconds and currency. For more information, see Internal dataset export.
Measuring compute with AIP¶
:::callout{theme="neutral"} If you have an enterprise contract with Palantir, contact your Palantir representative before proceeding with compute usage calculations. :::
| Model | Foundry cloud provider | Foundry region | Compute seconds per 10k input tokens | Compute seconds per 10k output tokens |
|---|---|---|---|---|
| Grok-2 ↗ | AWS | North America | 36 | 182 |
| AWS | EU / UK | 31 | 154 | |
| AWS | South America / APAC / Middle East | 25 | 125 | |
| Grok-2-Vision ↗ | AWS | North America | 36 | 182 |
| AWS | EU / UK | 31 | 154 | |
| AWS | South America / APAC / Middle East | 25 | 125 | |
| Grok-3 ↗ | AWS | North America | 55 | 273 |
| AWS | EU / UK | 46 | 231 | |
| AWS | South America / APAC / Middle East | 38 | 188 | |
| Grok-3-Mini-Reasoning ↗ | AWS | North America | 5.5 | 9.1 |
| AWS | EU / UK | 4.6 | 7.7 | |
| AWS | South America / APAC / Middle East | 3.8 | 6.3 | |
| Grok-4 <= 128k tokens ↗ | AWS | North America | 54.5 | 272.7 |
| AWS | EU / UK | 46.2 | 230.8 | |
| AWS | South America / APAC / Middle East | 37.5 | 187.5 | |
| Grok-4 > 128k tokens ↗ | AWS | North America | 109.1 | 545.5 |
| AWS | EU / UK | 92.3 | 461.5 | |
| AWS | South America / APAC / Middle East | 75.0 | 375.0 | |
| Grok-4 Fast Reasoning <= 128k tokens ↗ | AWS | North America | 3.6 | 9.1 |
| AWS | EU / UK | 3.1 | 7.7 | |
| AWS | South America / APAC / Middle East | 2.5 | 6.3 | |
| Grok-4 Fast Reasoning > 128k tokens ↗ | AWS | North America | 7.3 | 18.2 |
| AWS | EU / UK | 6.2 | 15.4 | |
| AWS | South America / APAC / Middle East | 5.0 | 12.5 | |
| Grok-4 Fast Non-Reasoning <= 128k tokens ↗ | AWS | North America | 3.6 | 9.1 |
| AWS | EU / UK | 3.1 | 7.7 | |
| AWS | South America / APAC / Middle East | 2.5 | 6.3 | |
| Grok-4 Fast Non-Reasoning > 128k tokens ↗ | AWS | North America | 7.3 | 18.2 |
| AWS | EU / UK | 6.2 | 15.4 | |
| AWS | South America / APAC / Middle East | 5.0 | 12.5 | |
| Grok Code Fast 1 ↗ | AWS | North America | 3.6 | 27.3 |
| AWS | EU / UK | 3.1 | 23.1 | |
| AWS | South America / APAC / Middle East | 2.5 | 18.8 | |
| Grok-4.1 Fast Non-Reasoning ↗ | AWS | North America | 3.6 | 9.1 |
| AWS | EU / UK | 3.1 | 7.7 | |
| AWS | South America / APAC / Middle East | 2.5 | 6.3 | |
| Grok-4.1 Fast Reasoning ↗ | AWS | North America | 3.6 | 9.1 |
| AWS | EU / UK | 3.1 | 7.7 | |
| AWS | South America / APAC / Middle East | 2.5 | 6.3 | |
| Grok-4.20 Reasoning <= 200k tokens ↗ | AWS | North America | 36.4 | 109.1 |
| AWS | EU / UK | 30.8 | 92.3 | |
| AWS | South America / APAC / Middle East | 25.0 | 75.0 | |
| Grok-4.20 Reasoning > 200k tokens ↗ | AWS | North America | 72.7 | 218.2 |
| AWS | EU / UK | 61.5 | 184.6 | |
| AWS | South America / APAC / Middle East | 50.0 | 150.0 | |
| Grok-4.20 Non-Reasoning <= 200k tokens ↗ | AWS | North America | 36.4 | 109.1 |
| AWS | EU / UK | 30.8 | 92.3 | |
| AWS | South America / APAC / Middle East | 25.0 | 75.0 | |
| Grok-4.20 Non-Reasoning > 200k tokens ↗ | AWS | North America | 72.7 | 218.2 |
| AWS | EU / UK | 61.5 | 184.6 | |
| AWS | South America / APAC / Middle East | 50.0 | 150.0 | |
| GPT-4.5 ↗ | AWS | North America | 1159.1 | 2318.2 |
| AWS | EU / UK | 980.8 | 1961.5 | |
| AWS | South America / APAC / Middle East | 796.9 | 1593.8 | |
| GPT-4o ↗ | AWS | North America | 43 | 172 |
| AWS | EU / UK | 36 | 145 | |
| AWS | South America / APAC / Middle East | 30 | 118 | |
| GPT-4o mini ↗ | AWS | North America | 2.6 | 10.3 |
| AWS | EU / UK | 2.2 | 8.7 | |
| AWS | South America / APAC / Middle East | 1.8 | 7.1 | |
| GPT-4.1 ↗ | AWS | North America | 31 | 124 |
| AWS | EU / UK | 26 | 105 | |
| AWS | South America / APAC / Middle East | 21 | 85 | |
| GPT-4.1-mini ↗ | AWS | North America | 6.2 | 24.7 |
| AWS | EU / UK | 5.2 | 20.9 | |
| AWS | South America / APAC / Middle East | 4.3 | 17 | |
| GPT-4.1-nano ↗ | AWS | North America | 1.5 | 6.2 |
| AWS | EU / UK | 1.3 | 5.2 | |
| AWS | South America / APAC / Middle East | 1.1 | 4.3 | |
| GPT-5 ↗ | AWS | North America | 20.5 | 163.6 |
| AWS | EU / UK | 17.3 | 138.5 | |
| AWS | South America / APAC / Middle East | 14.1 | 112.5 | |
| GPT-5-mini ↗ | AWS | North America | 4.1 | 32.7 |
| AWS | EU / UK | 3.5 | 27.7 | |
| AWS | South America / APAC / Middle East | 2.8 | 22.5 | |
| GPT-5-nano ↗ | AWS | North America | 0.82 | 6.5 |
| AWS | EU / UK | 0.69 | 5.5 | |
| AWS | South America / APAC / Middle East | 0.56 | 4.5 | |
| GPT-5-pro ↗ | AWS | North America | 231.8 | 1854.5 |
| AWS | EU / UK | 196.2 | 1569.2 | |
| AWS | South America / APAC / Middle East | 159.4 | 1275.0 | |
| GPT-OSS-20B ↗ | AWS | North America | 1.1 | 4.9 |
| AWS | EU / UK | 1.0 | 4.2 | |
| AWS | South America / APAC / Middle East | 0.79 | 3.4 | |
| GPT-OSS-120B ↗ | AWS | North America | 2.5 | 9.8 |
| AWS | EU / UK | 2.1 | 8.3 | |
| AWS | South America / APAC / Middle East | 1.7 | 6.8 | |
| GPT-5 Codex ↗ | AWS | North America | 20.5 | 163.6 |
| AWS | EU / UK | 17.3 | 138.5 | |
| AWS | South America / APAC / Middle East | 14.1 | 112.5 | |
| GPT-5.1 Codex Mini ↗ | AWS | North America | 5.5 | 36.4 |
| AWS | EU / UK | 4.6 | 30.8 | |
| AWS | South America / APAC / Middle East | 3.8 | 25 | |
| GPT-5.1 Codex ↗ | AWS | North America | 23.6 | 181.8 |
| AWS | EU / UK | 20 | 153.8 | |
| AWS | South America / APAC / Middle East | 16.3 | 125 | |
| GPT-5.1 ↗ | AWS | North America | 23.6 | 181.8 |
| AWS | EU / UK | 20 | 153.8 | |
| AWS | South America / APAC / Middle East | 16.3 | 125 | |
| GPT-5.1 Codex Max ↗ | AWS | North America | 22.7 | 181.8 |
| AWS | EU / UK | 19.2 | 153.8 | |
| AWS | South America / APAC / Middle East | 15.6 | 125.0 | |
| GPT-5.2 ↗ | AWS | North America | 31.8 | 254.5 |
| AWS | EU / UK | 26.9 | 215.4 | |
| AWS | South America / APAC / Middle East | 21.9 | 175.0 | |
| GPT-5.2 Codex ↗ | AWS | North America | 32.7 | 254.5 |
| AWS | EU / UK | 27.7 | 215.4 | |
| AWS | South America / APAC / Middle East | 22.5 | 175 | |
| GPT-5.2 Pro ↗ | AWS | North America | 381.8 | 3054.5 |
| AWS | EU / UK | 323.1 | 2584.6 | |
| AWS | South America / APAC / Middle East | 262.5 | 2100.0 | |
| GPT-5.3 Codex ↗ | AWS | North America | 31.8 | 254.5 |
| AWS | EU / UK | 26.9 | 215.4 | |
| AWS | South America / APAC / Middle East | 21.9 | 175.0 | |
| GPT-5.4 <= 272k tokens ↗ | AWS | North America | 45.5 | 272.7 |
| AWS | EU / UK | 38.5 | 230.8 | |
| AWS | South America / APAC / Middle East | 31.3 | 187.5 | |
| GPT-5.4 > 272k tokens ↗ | AWS | North America | 90.9 | 409.1 |
| AWS | EU / UK | 76.9 | 346.2 | |
| AWS | South America / APAC / Middle East | 62.5 | 281.3 | |
| GPT-5.4 Pro <= 272k tokens ↗ | AWS | North America | 545.5 | 3272.7 |
| AWS | EU / UK | 461.5 | 2769.2 | |
| AWS | South America / APAC / Middle East | 375.0 | 2250.0 | |
| GPT-5.4 Pro > 272k tokens ↗ | AWS | North America | 1090.9 | 4909.1 |
| AWS | EU / UK | 923.1 | 4153.8 | |
| AWS | South America / APAC / Middle East | 750.0 | 3375.0 | |
| GPT-5.4-mini ↗ | AWS | North America | 13.6 | 81.8 |
| AWS | EU / UK | 11.5 | 69.2 | |
| AWS | South America / APAC / Middle East | 9.4 | 56.3 | |
| GPT-5.4-nano ↗ | AWS | North America | 3.6 | 22.7 |
| AWS | EU / UK | 3.1 | 19.2 | |
| AWS | South America / APAC / Middle East | 2.5 | 15.6 | |
| GPT-5.5 <= 272k tokens ↗ | AWS | North America | 81.8 | 490.9 |
| AWS | EU / UK | 69.2 | 415.4 | |
| AWS | South America / APAC / Middle East | 56.3 | 337.5 | |
| GPT-5.5 > 272k tokens ↗ | AWS | North America | 163.6 | 736.4 |
| AWS | EU / UK | 138.5 | 623.1 | |
| AWS | South America / APAC / Middle East | 112.5 | 506.3 | |
| GPT Realtime ↗ | AWS | North America | 72.7 | 290.9 |
| AWS | EU / UK | 61.5 | 246.2 | |
| AWS | South America / APAC / Middle East | 50 | 200 | |
| GPT Realtime 1.5 ↗ | AWS | North America | 72.7 | 290.9 |
| AWS | EU / UK | 61.5 | 246.2 | |
| AWS | South America / APAC / Middle East | 50.0 | 200.0 | |
| o1 ↗ | AWS | North America | 232 | 927 |
| AWS | EU / UK | 196 | 785 | |
| AWS | South America / APAC / Middle East | 159 | 638 | |
| o1-mini ↗ | AWS | North America | 17 | 68 |
| AWS | EU / UK | 14 | 58 | |
| AWS | South America / APAC / Middle East | 12 | 47 | |
| o3 ↗ | AWS | North America | 31 | 124 |
| AWS | EU / UK | 26 | 105 | |
| AWS | South America / APAC / Middle East | 21 | 85 | |
| o3-mini ↗ | AWS | North America | 17 | 68 |
| AWS | EU / UK | 14 | 58 | |
| AWS | South America / APAC / Middle East | 12 | 47 | |
| o3-pro ↗ | AWS | North America | 345.5 | 1381.8 |
| AWS | EU / UK | 292.3 | 1169.2 | |
| AWS | South America / APAC / Middle East | 237.5 | 950.0 | |
| o4-mini ↗ | AWS | North America | 17 | 68 |
| AWS | EU / UK | 14 | 58 | |
| AWS | South America / APAC / Middle East | 12 | 47 | |
ada embedding ↗ |
AWS | North America | 1.68 | N/A |
| AWS | EU / UK | 1.42 | N/A | |
| AWS | South America / APAC / Middle East | 1.16 | N/A | |
| text-embedding-3-large ↗ | AWS | North America | 2.24 | N/A |
| AWS | EU / UK | 1.89 | N/A | |
| AWS | South America / APAC / Middle East | 1.54 | N/A | |
| text-embedding-3-small ↗ | AWS | North America | 0.34 | N/A |
| AWS | EU / UK | 0.29 | N/A | |
| AWS | South America / APAC / Middle East | 0.24 | N/A | |
| OpenAI Text Embedding 3 Large ↗ | AWS | North America | 2.2 | N/A |
| AWS | EU / UK | 1.9 | N/A | |
| AWS | South America / APAC / Middle East | 1.5 | N/A | |
| OpenAI Text Embedding 3 Small ↗ | AWS | North America | 0.3 | N/A |
| AWS | EU / UK | 0.3 | N/A | |
| AWS | South America / APAC / Middle East | 0.2 | N/A | |
| OpenAI Text Embedding Ada 002 ↗ | AWS | North America | 1.7 | N/A |
| AWS | EU / UK | 1.4 | N/A | |
| AWS | South America / APAC / Middle East | 1.2 | N/A | |
| Anthropic Claude 3 ↗ | AWS | North America | 52 | 258 |
| AWS | EU / UK | 44 | 218 | |
| AWS | South America / APAC / Middle East | 35 | 177 | |
| Anthropic Claude 3 Haiku ↗ | AWS | North America | 4.3 | 21.5 |
| AWS | EU / UK | 3.6 | 18.2 | |
| AWS | South America / APAC / Middle East | 3.0 | 14.8 | |
| Anthropic Claude 3.5 Haiku ↗ | AWS | North America | 12 | 62 |
| AWS | EU / UK | 10 | 52 | |
| AWS | South America / APAC / Middle East | 9 | 43 | |
| Anthropic Claude 4.5 Haiku ↗ | AWS | North America | 17.3 | 86.4 |
| AWS | EU / UK | 14.6 | 73.1 | |
| AWS | South America / APAC / Middle East | 11.9 | 59.4 | |
| Anthropic Claude 3.5 Sonnet ↗ | AWS | North America | 52 | 258 |
| AWS | EU / UK | 44 | 218 | |
| AWS | South America / APAC / Middle East | 35 | 177 | |
| Anthropic Claude 3.5 Sonnet v2 ↗ | AWS | North America | 46 | 232 |
| AWS | EU / UK | 39 | 196 | |
| AWS | South America / APAC / Middle East | 32 | 159 | |
| Anthropic Claude 4 Sonnet ↗ | AWS | North America | 46.4 | 231.8 |
| AWS | EU / UK | 39.2 | 196.2 | |
| AWS | South America / APAC / Middle East | 31.9 | 159.4 | |
| Anthropic Claude 4.5 Sonnet ↗ | AWS | North America | 51.8 | 259.1 |
| AWS | EU / UK | 43.8 | 219.2 | |
| AWS | South America / APAC / Middle East | 35.6 | 178.1 | |
| Anthropic Claude 4.6 Sonnet ↗ | AWS | North America | 54.5 | 272.7 |
| AWS | EU / UK | 46.2 | 230.8 | |
| AWS | South America / APAC / Middle East | 37.5 | 187.5 | |
| Anthropic Claude 4 Opus ↗ | AWS | North America | 232 | 1159 |
| AWS | EU / UK | 196 | 981 | |
| AWS | South America / APAC / Middle East | 159 | 797 | |
| Anthropic Claude 4.1 Opus ↗ | AWS | North America | 259 | 1295 |
| AWS | EU / UK | 219 | 1096 | |
| AWS | South America / APAC / Middle East | 178 | 891 | |
| Anthropic Claude 4.5 Opus ↗ | AWS | North America | 90.9 | 454.5 |
| AWS | EU / UK | 76.9 | 384.6 | |
| AWS | South America / APAC / Middle East | 62.5 | 312.5 | |
| Anthropic Claude 4.6 Opus ↗ | AWS | North America | 90.9 | 454.5 |
| AWS | EU / UK | 76.9 | 384.6 | |
| AWS | South America / APAC / Middle East | 62.5 | 312.5 | |
| Anthropic Claude 4.7 Opus ↗ | AWS | North America | 90.9 | 454.5 |
| AWS | EU / UK | 76.9 | 384.6 | |
| AWS | South America / APAC / Middle East | 62.5 | 312.5 | |
| Anthropic Claude 4.8 Opus ↗ | AWS | North America | 90.9 | 454.5 |
| AWS | EU / UK | 76.9 | 384.6 | |
| AWS | South America / APAC / Middle East | 62.5 | 312.5 | |
| Mistral Small 24B ↗ | AWS | North America | 158 | 525 |
| AWS | EU / UK | 133 | 444 | |
| AWS | South America / APAC / Middle East | 108 | 361 | |
| Mistral Small 24B Instruct ↗ | AWS | North America | 157.5 | 525 |
| AWS | EU / UK | 133.3 | 444.2 | |
| AWS | South America / APAC / Middle East | 108.3 | 360.9 | |
| Llama 3.1_8B ↗ | AWS | North America | 158 | 525 |
| AWS | EU / UK | 133 | 444 | |
| AWS | South America / APAC / Middle East | 108 | 361 | |
| Llama 3.3_70B ↗ | AWS | North America | 158 | 525 |
| AWS | EU / UK | 133 | 444 | |
| AWS | South America / APAC / Middle East | 108 | 361 | |
| Llama 3.3 70B Instruct ↗ | AWS | North America | 157.5 | 525 |
| AWS | EU / UK | 133.3 | 444.2 | |
| AWS | South America / APAC / Middle East | 108.3 | 360.9 | |
| Llama 4 Scout_17B 16E Instruct ↗ | AWS | North America | 1.5 | 5.7 |
| AWS | EU / UK | 1.2 | 4.8 | |
| AWS | South America / APAC / Middle East | 1.0 | 3.9 | |
| Llama 4 Maverick_17B 128E Instruct ↗ | AWS | North America | 2.1 | 8.4 |
| AWS | EU / UK | 1.8 | 7.1 | |
| AWS | South America / APAC / Middle East | 1.4 | 5.8 | |
| Nemotron 3 Nano 30B ↗ | AWS | North America | 1.1 | 4.4 |
| AWS | EU / UK | 0.9 | 3.7 | |
| AWS | South America / APAC / Middle East | 0.8 | 3.0 | |
| Nemotron 3 Super 120B ↗ | AWS | North America | 2.7 | 11.8 |
| AWS | EU / UK | 2.3 | 10.0 | |
| AWS | South America / APAC / Middle East | 1.9 | 8.1 | |
| Snowflake Arctic Embed ↗ | AWS | North America | 38 | 38 |
| AWS | EU / UK | 32 | 32 | |
| AWS | South America / APAC / Middle East | 26 | 26 | |
| Snowflake Arctic Embed M ↗ | AWS | North America | 38.2 | 38.2 |
| AWS | EU / UK | 32.3 | 32.3 | |
| AWS | South America / APAC / Middle East | 26.2 | 26.2 | |
| Gemini 1.5 Flash ↗ | AWS | North America | 1.3 | 5.2 |
| AWS | EU / UK | 1.1 | 4.4 | |
| AWS | South America / APAC / Middle East | 0.9 | 3.5 | |
| Gemini 1.5 Pro ↗ | AWS | North America | 21 | 86 |
| AWS | EU / UK | 18 | 73 | |
| AWS | South America / APAC / Middle East | 15 | 59 | |
| Gemini 2.0 Flash ↗ | AWS | North America | 1.5 | 6.2 |
| AWS | EU / UK | 1.3 | 5.2 | |
| AWS | South America / APAC / Middle East | 1.1 | 4.3 | |
| Gemini 2.5 Flash Lite ↗ | AWS | North America | 1.7 | 6.9 |
| AWS | EU / UK | 1.5 | 5.8 | |
| AWS | South America / APAC / Middle East | 1.2 | 4.8 | |
| Gemini 2.5 Flash ↗ | AWS | North America | 5.2 | 43.2 |
| AWS | EU / UK | 4.4 | 36.5 | |
| AWS | South America / APAC / Middle East | 3.6 | 29.7 | |
| Gemini 2.5 Pro <= 200k tokens ↗ | AWS | North America | 21.6 | 172.7 |
| AWS | EU / UK | 18.3 | 146.2 | |
| AWS | South America / APAC / Middle East | 14.8 | 118.8 | |
| Gemini 2.5 Pro > 200k tokens ↗ | AWS | North America | 43.2 | 259.1 |
| AWS | EU / UK | 36.5 | 219.1 | |
| AWS | South America / APAC / Middle East | 29.7 | 178.1 | |
| Gemini 3 Flash ↗ | AWS | North America | 9.1 | 54.5 |
| AWS | EU / UK | 7.7 | 46.2 | |
| AWS | South America / APAC / Middle East | 6.3 | 37.5 | |
| Gemini 3 Pro <= 200k tokens ↗ | AWS | North America | 34.5 | 207.3 |
| AWS | EU / UK | 29.2 | 175.4 | |
| AWS | South America / APAC / Middle East | 23.8 | 142.5 | |
| Gemini 3 Pro > 200k tokens ↗ | AWS | North America | 69.1 | 310.9 |
| AWS | EU / UK | 58.5 | 263.1 | |
| AWS | South America / APAC / Middle East | 47.5 | 213.8 | |
| Gemini 3.1 Pro <= 200k tokens ↗ | AWS | North America | 34.5 | 207.3 |
| AWS | EU / UK | 29.2 | 175.4 | |
| AWS | South America / APAC / Middle East | 23.8 | 142.5 | |
| Gemini 3.1 Pro > 200k tokens ↗ | AWS | North America | 69.1 | 310.9 |
| AWS | EU / UK | 58.5 | 263.1 | |
| AWS | South America / APAC / Middle East | 47.5 | 213.8 | |
| Gemini 3.1 Flash Lite ↗ | AWS | North America | 4.5 | 27.3 |
| AWS | EU / UK | 3.8 | 23.1 | |
| AWS | South America / APAC / Middle East | 3.1 | 18.8 | |
| Gemini Embedding 2 Text ↗ | AWS | North America | 3.6 | N/A |
| AWS | EU / UK | 3.1 | N/A | |
| AWS | South America / APAC / Middle East | 2.5 | N/A | |
| Document Information Extraction | AWS | North America | 182 | N/A |
| AWS | EU / UK | 154 | N/A | |
| AWS | South America / APAC / Middle East | 125 | N/A |
中文翻译¶
AIP 计算用量¶
AIP 计算用量涉及大型语言模型（LLM，large language models）。从根本上说，LLM 接收文本作为输入，并以文本作为输出进行响应。输入和输出的文本量以 token 为单位进行衡量。LLM 的计算用量 以每一定数量 token 的计算秒数（compute-seconds）来衡量。不同模型的计算用量费率可能不同，详见下文。
AIP 中的 Token¶
Token 是 LLM 用于处理和理解输入文本的基本单位。根据语言和具体模型的不同，一个 token 可以短至单个字符，也可以长至整个单词。
重要的是，token 与单词并非一一对应。例如，常见单词可能是一个 token，但较长或不常见的单词可能会被拆分为多个 token。甚至标点符号和空格也可能被视为 token。
不同的模型提供商对 token 的定义各不相同；例如，OpenAI ↗ 和 Anthropic ↗。平均而言，一个 token 大约包含 4 个字符，一个字符可以是单个字母或标点符号。
在 AIP 中，向 LLM 发送提示（prompt）并从 LLM 接收提示的应用程序会消耗 token。每个提示和响应都由可衡量的 token 数量组成。这些 token 可以发送给多个 LLM 提供商；由于提供商之间的差异，这些 token 会被转换为计算秒数，以匹配底层模型提供商的价格。
所有提供 LLM 支持功能的应用程序在使用时都会消耗 token。请参阅以下列表，了解当您与这些应用程序的 LLM 支持功能交互时可能消耗 token 的应用程序。
- AIP Assist
- AIP Logic
- AIP Error Enhancer
- AIP Code Assist
- AIP Analyst
- AI FDE
- Workshop LLM 支持工具
- Quiver LLM 支持工具
- Pipeline Builder LLM 支持工具
- 直接调用语言模型服务（包括 Python 和 TypeScript 库）
AIP 将文本直接路由到后端 LLM，这些 LLM 自行执行分词（tokenization）。文本的大小将决定后端模型用于提供响应所消耗的计算量。
以下示例句子被发送到 GPT-4o 模型。
AIP incorporates all of Palantir's advanced security measures for the protection of sensitive data in compliance with industry regulations.
该句子包含 140 个字符，将按以下方式进行分词，每个 token 之间用 | 字符分隔。请注意，token 并不总是等同于一个单词；有些单词会被拆分为多个 token，如下例中的 AIP 和 Palantir。
A|IP| incorporates| all| of| Pal|ant|ir|'s| advanced| security| measures| for| the| protection| of| sensitive| data| in| compliance| with| industry| regulations|.
该句子包含 24 个 token，将消耗以下数量的计算秒数：
计算秒数 = 24 个 token * 43 计算秒数 / 10,000 个 token
计算秒数 = 24 * 43 / 10,000
计算秒数 = 0.1032
上述句子中的 token 数量和字符数已通过 OpenAI 的 Tokenizer 功能 ↗ 验证。
了解 AIP 计算用量的驱动因素¶
由 LLM token 产生的计算秒数用量直接归属于请求该用量的单个应用程序资源。例如，如果您使用 AIP 自动解释 Pipeline Builder 中的管道，那么 LLM 生成该解释所使用的计算秒数将归属于该特定管道。这在平台中都是如此；牢记这一点将帮助您跟踪 token 的使用位置。
在某些情况下，计算用量无法归属于平台中的单个资源；例如 AIP Assist 和 Error Explainer 等。当用量无法归属于单个资源时，token 将归属于发起 token 使用的用户文件夹。
我们建议您注意代表您发送给 LLM 的 token。通常，您在使用 LLM 时包含的信息越多，消耗的计算秒数就越多。例如，以下场景描述了使用计算秒数的不同方式。
- 在 Pipeline Builder 中，您可以要求 AIP 解释您的转换节点；所选节点的数量会影响 LLM 生成响应所使用的 token 数量，从而影响计算秒数的用量。这是因为随着节点数量的增加，LLM 必须处理的关于这些节点配置的文本量也会增加。
- 在 AIP Assist 中，要求 LLM 生成大块代码需要更多的输出 token。较短的响应使用较少的 token，因此计算量也更少。
- 在 AIP Logic 中，在提示中发送大量文本需要更多的 token，因此需要更多的计算秒数。
导出 AIP Token 用量数据¶
要详细分析您的注册（enrollment）的 LLM 用量，您可以从控制面板的内部数据集导出部分导出 AIP Token Usage 数据集。该数据集提供按模型和资源划分的每日 token 消耗明细，以及相应的计算秒数和货币用量。有关更多信息，请参阅内部数据集导出。
使用 AIP 衡量计算¶
:::callout{theme="neutral"} 如果您与 Palantir 签订了企业合同，请在进行计算用量计算之前联系您的 Palantir 代表。 :::
| 模型 | Foundry 云提供商 | Foundry 区域 | 每 10k 输入 token 的计算秒数 | 每 10k 输出 token 的计算秒数 |
|---|---|---|---|---|
| Grok-2 ↗ | AWS | 北美 | 36 | 182 |
| AWS | 欧盟 / 英国 | 31 | 154 | |
| AWS | 南美 / 亚太 / 中东 | 25 | 125 | |
| Grok-2-Vision ↗ | AWS | 北美 | 36 | 182 |
| AWS | 欧盟 / 英国 | 31 | 154 | |
| AWS | 南美 / 亚太 / 中东 | 25 | 125 | |
| Grok-3 ↗ | AWS | 北美 | 55 | 273 |
| AWS | 欧盟 / 英国 | 46 | 231 | |
| AWS | 南美 / 亚太 / 中东 | 38 | 188 | |
| Grok-3-Mini-Reasoning ↗ | AWS | 北美 | 5.5 | 9.1 |
| AWS | 欧盟 / 英国 | 4.6 | 7.7 | |
| AWS | 南美 / 亚太 / 中东 | 3.8 | 6.3 | |
| Grok-4 <= 128k tokens ↗ | AWS | 北美 | 54.5 | 272.7 |
| AWS | 欧盟 / 英国 | 46.2 | 230.8 | |
| AWS | 南美 / 亚太 / 中东 | 37.5 | 187.5 | |
| Grok-4 > 128k tokens ↗ | AWS | 北美 | 109.1 | 545.5 |
| AWS | 欧盟 / 英国 | 92.3 | 461.5 | |
| AWS | 南美 / 亚太 / 中东 | 75.0 | 375.0 | |
| Grok-4 Fast Reasoning <= 128k tokens ↗ | AWS | 北美 | 3.6 | 9.1 |
| AWS | 欧盟 / 英国 | 3.1 | 7.7 | |
| AWS | 南美 / 亚太 / 中东 | 2.5 | 6.3 | |
| Grok-4 Fast Reasoning > 128k tokens ↗ | AWS | 北美 | 7.3 | 18.2 |
| AWS | 欧盟 / 英国 | 6.2 | 15.4 | |
| AWS | 南美 / 亚太 / 中东 | 5.0 | 12.5 | |
| Grok-4 Fast Non-Reasoning <= 128k tokens ↗ | AWS | 北美 | 3.6 | 9.1 |
| AWS | 欧盟 / 英国 | 3.1 | 7.7 | |
| AWS | 南美 / 亚太 / 中东 | 2.5 | 6.3 | |
| Grok-4 Fast Non-Reasoning > 128k tokens ↗ | AWS | 北美 | 7.3 | 18.2 |
| AWS | 欧盟 / 英国 | 6.2 | 15.4 | |
| AWS | 南美 / 亚太 / 中东 | 5.0 | 12.5 | |
| Grok Code Fast 1 ↗ | AWS | 北美 | 3.6 | 27.3 |
| AWS | 欧盟 / 英国 | 3.1 | 23.1 | |
| AWS | 南美 / 亚太 / 中东 | 2.5 | 18.8 | |
| Grok-4.1 Fast Non-Reasoning ↗ | AWS | 北美 | 3.6 | 9.1 |
| AWS | 欧盟 / 英国 | 3.1 | 7.7 | |
| AWS | 南美 / 亚太 / 中东 | 2.5 | 6.3 | |
| Grok-4.1 Fast Reasoning ↗ | AWS | 北美 | 3.6 | 9.1 |
| AWS | 欧盟 / 英国 | 3.1 | 7.7 | |
| AWS | 南美 / 亚太 / 中东 | 2.5 | 6.3 | |
| Grok-4.20 Reasoning <= 200k tokens ↗ | AWS | 北美 | 36.4 | 109.1 |
| AWS | 欧盟 / 英国 | 30.8 | 92.3 | |
| AWS | 南美 / 亚太 / 中东 | 25.0 | 75.0 | |
| Grok-4.20 Reasoning > 200k tokens ↗ | AWS | 北美 | 72.7 | 218.2 |
| AWS | 欧盟 / 英国 | 61.5 | 184.6 | |
| AWS | 南美 / 亚太 / 中东 | 50.0 | 150.0 | |
| Grok-4.20 Non-Reasoning <= 200k tokens ↗ | AWS | 北美 | 36.4 | 109.1 |
| AWS | 欧盟 / 英国 | 30.8 | 92.3 | |
| AWS | 南美 / 亚太 / 中东 | 25.0 | 75.0 | |
| Grok-4.20 Non-Reasoning > 200k tokens ↗ | AWS | 北美 | 72.7 | 218.2 |
| AWS | 欧盟 / 英国 | 61.5 | 184.6 | |
| AWS | 南美 / 亚太 / 中东 | 50.0 | 150.0 | |
| GPT-4.5 ↗ | AWS | 北美 | 1159.1 | 2318.2 |
| AWS | 欧盟 / 英国 | 980.8 | 1961.5 | |
| AWS | 南美 / 亚太 / 中东 | 796.9 | 1593.8 | |
| GPT-4o ↗ | AWS | 北美 | 43 | 172 |
| AWS | 欧盟 / 英国 | 36 | 145 | |
| AWS | 南美 / 亚太 / 中东 | 30 | 118 | |
| GPT-4o mini ↗ | AWS | 北美 | 2.6 | 10.3 |
| AWS | 欧盟 / 英国 | 2.2 | 8.7 | |
| AWS | 南美 / 亚太 / 中东 | 1.8 | 7.1 | |
| GPT-4.1 ↗ | AWS | 北美 | 31 | 124 |
| AWS | 欧盟 / 英国 | 26 | 105 | |
| AWS | 南美 / 亚太 / 中东 | 21 | 85 | |
| GPT-4.1-mini ↗ | AWS | 北美 | 6.2 | 24.7 |
| AWS | 欧盟 / 英国 | 5.2 | 20.9 | |
| AWS | 南美 / 亚太 / 中东 | 4.3 | 17 | |
| GPT-4.1-nano ↗ | AWS | 北美 | 1.5 | 6.2 |
| AWS | 欧盟 / 英国 | 1.3 | 5.2 | |
| AWS | 南美 / 亚太 / 中东 | 1.1 | 4.3 | |
| GPT-5 ↗ | AWS | 北美 | 20.5 | 163.6 |
| AWS | 欧盟 / 英国 | 17.3 | 138.5 | |
| AWS | 南美 / 亚太 / 中东 | 14.1 | 112.5 | |
| GPT-5-mini ↗ | AWS | 北美 | 4.1 | 32.7 |
| AWS | 欧盟 / 英国 | 3.5 | 27.7 | |
| AWS | 南美 / 亚太 / 中东 | 2.8 | 22.5 | |
| GPT-5-nano ↗ | AWS | 北美 | 0.82 | 6.5 |
| AWS | 欧盟 / 英国 | 0.69 | 5.5 | |
| AWS | 南美 / 亚太 / 中东 | 0.56 | 4.5 | |
| GPT-5-pro ↗ | AWS | 北美 | 231.8 | 1854.5 |
| AWS | 欧盟 / 英国 | 196.2 | 1569.2 | |
| AWS | 南美 / 亚太 / 中东 | 159.4 | 1275.0 | |
| GPT-OSS-20B ↗ | AWS | 北美 | 1.1 | 4.9 |
| AWS | 欧盟 / 英国 | 1.0 | 4.2 | |
| AWS | 南美 / 亚太 / 中东 | 0.79 | 3.4 | |
| GPT-OSS-120B ↗ | AWS | 北美 | 2.5 | 9.8 |
| AWS | 欧盟 / 英国 | 2.1 | 8.3 | |
| AWS | 南美 / 亚太 / 中东 | 1.7 | 6.8 | |
| GPT-5 Codex ↗ | AWS | 北美 | 20.5 | 163.6 |
| AWS | 欧盟 / 英国 | 17.3 | 138.5 | |
| AWS | 南美 / 亚太 / 中东 | 14.1 | 112.5 | |
| GPT-5.1 Codex Mini ↗ | AWS | 北美 | 5.5 | 36.4 |
| AWS | 欧盟 / 英国 | 4.6 | 30.8 | |
| AWS | 南美 / 亚太 / 中东 | 3.8 | 25 | |
| GPT-5.1 Codex ↗ | AWS | 北美 | 23.6 | 181.8 |
| AWS | 欧盟 / 英国 | 20 | 153.8 | |
| AWS | 南美 / 亚太 / 中东 | 16.3 | 125 | |
| GPT-5.1 ↗ | AWS | 北美 | 23.6 | 181.8 |
| AWS | 欧盟 / 英国 | 20 | 153.8 | |
| AWS | 南美 / 亚太 / 中东 | 16.3 | 125 | |
| GPT-5.1 Codex Max ↗ | AWS | 北美 | 22.7 | 181.8 |
| AWS | 欧盟 / 英国 | 19.2 | 153.8 | |
| AWS | 南美 / 亚太 / 中东 | 15.6 | 125.0 | |
| GPT-5.2 ↗ | AWS | 北美 | 31.8 | 254.5 |
| AWS | 欧盟 / 英国 | 26.9 | 215.4 | |
| AWS | 南美 / 亚太 / 中东 | 21.9 | 175.0 | |
| GPT-5.2 Codex ↗ | AWS | 北美 | 32.7 | 254.5 |
| AWS | 欧盟 / 英国 | 27.7 | 215.4 | |
| AWS | 南美 / 亚太 / 中东 | 22.5 | 175 | |
| GPT-5.2 Pro ↗ | AWS | 北美 | 381.8 | 3054.5 |
| AWS | 欧盟 / 英国 | 323.1 | 2584.6 | |
| AWS | 南美 / 亚太 / 中东 | 262.5 | 2100.0 | |
| GPT-5.3 Codex ↗ | AWS | 北美 | 31.8 | 254.5 |
| AWS | 欧盟 / 英国 | 26.9 | 215.4 | |
| AWS | 南美 / 亚太 / 中东 | 21.9 | 175.0 | |
| GPT-5.4 <= 272k tokens ↗ | AWS | 北美 | 45.5 | 272.7 |
| AWS | 欧盟 / 英国 | 38.5 | 230.8 | |
| AWS | 南美 / 亚太 / 中东 | 31.3 | 187.5 | |
| GPT-5.4 > 272k tokens ↗ | AWS | 北美 | 90.9 | 409.1 |
| AWS | 欧盟 / 英国 | 76.9 | 346.2 | |
| AWS | 南美 / 亚太 / 中东 | 62.5 | 281.3 | |
| GPT-5.4 Pro <= 272k tokens ↗ | AWS | 北美 | 545.5 | 3272.7 |
| AWS | 欧盟 / 英国 | 461.5 | 2769.2 | |
| AWS | 南美 / 亚太 / 中东 | 375.0 | 2250.0 | |
| GPT-5.4 Pro > 272k tokens ↗ | AWS | 北美 | 1090.9 | 4909.1 |
| AWS | 欧盟 / 英国 | 923.1 | 4153.8 | |
| AWS | 南美 / 亚太 / 中东 | 750.0 | 3375.0 | |
| GPT-5.4-mini ↗ | AWS | 北美 | 13.6 | 81.8 |
| AWS | 欧盟 / 英国 | 11.5 | 69.2 | |
| AWS | 南美 / 亚太 / 中东 | 9.4 | 56.3 | |
| GPT-5.4-nano ↗ | AWS | 北美 | 3.6 | 22.7 |
| AWS | 欧盟 / 英国 | 3.1 | 19.2 | |
| AWS | 南美 / 亚太 / 中东 | 2.5 | 15.6 | |
| GPT-5.5 <= 272k tokens ↗ | AWS | 北美 | 81.8 | 490.9 |
| AWS | 欧盟 / 英国 | 69.2 | 415.4 | |
| AWS | 南美 / 亚太 / 中东 | 56.3 | 337.5 | |
| GPT-5.5 > 272k tokens ↗ | AWS | 北美 | 163.6 | 736.4 |
| AWS | 欧盟 / 英国 | 138.5 | 623.1 | |
| AWS | 南美 / 亚太 / 中东 | 112.5 | 506.3 | |
| GPT Realtime ↗ | AWS | 北美 | 72.7 | 290.9 |
| AWS | 欧盟 / 英国 | 61.5 | 246.2 | |
| AWS | 南美 / 亚太 / 中东 | 50 | 200 | |
| GPT Realtime 1.5 ↗ | AWS | 北美 | 72.7 | 290.9 |
| AWS | 欧盟 / 英国 | 61.5 | 246.2 | |
| AWS | 南美 / 亚太 / 中东 | 50.0 | 200.0 | |
| o1 ↗ | AWS | 北美 | 232 | 927 |
| AWS | 欧盟 / 英国 | 196 | 785 | |
| AWS | 南美 / 亚太 / 中东 | 159 | 638 | |
| o1-mini ↗ | AWS | 北美 | 17 | 68 |
| AWS | 欧盟 / 英国 | 14 | 58 | |
| AWS | 南美 / 亚太 / 中东 | 12 | 47 | |
| o3 ↗ | AWS | 北美 | 31 | 124 |
| AWS | 欧盟 / 英国 | 26 | 105 | |
| AWS | 南美 / 亚太 / 中东 | 21 | 85 | |
| o3-mini ↗ | AWS | 北美 | 17 | 68 |
| AWS | 欧盟 / 英国 | 14 | 58 | |
| AWS | 南美 / 亚太 / 中东 | 12 | 47 | |
| o3-pro ↗ | AWS | 北美 | 345.5 | 1381.8 |
| AWS | 欧盟 / 英国 | 292.3 | 1169.2 | |
| AWS | 南美 / 亚太 / 中东 | 237.5 | 950.0 | |
| o4-mini ↗ | AWS | 北美 | 17 | 68 |
| AWS | 欧盟 / 英国 | 14 | 58 | |
| AWS | 南美 / 亚太 / 中东 | 12 | 47 | |
ada embedding ↗ |
AWS | 北美 | 1.68 | 不适用 |
| AWS | 欧盟 / 英国 | 1.42 | 不适用 | |
| AWS | 南美 / 亚太 / 中东 | 1.16 | 不适用 | |
| text-embedding-3-large ↗ | AWS | 北美 | 2.24 | 不适用 |
| AWS | 欧盟 / 英国 | 1.89 | 不适用 | |
| AWS | 南美 / 亚太 / 中东 | 1.54 | 不适用 | |
| text-embedding-3-small ↗ | AWS | 北美 | 0.34 | 不适用 |
| AWS | 欧盟 / 英国 | 0.29 | 不适用 | |
| AWS | 南美 / 亚太 / 中东 | 0.24 | 不适用 | |
| OpenAI Text Embedding 3 Large ↗ | AWS | 北美 | 2.2 | 不适用 |
| AWS | 欧盟 / 英国 | 1.9 | 不适用 | |
| AWS | 南美 / 亚太 / 中东 | 1.5 | 不适用 | |
| OpenAI Text Embedding 3 Small ↗ | AWS | 北美 | 0.3 | 不适用 |
| AWS | 欧盟 / 英国 | 0.3 | 不适用 | |
| AWS | 南美 / 亚太 / 中东 | 0.2 | 不适用 | |
| OpenAI Text Embedding Ada 002 ↗ | AWS | 北美 | 1.7 | 不适用 |
| AWS | 欧盟 / 英国 | 1.4 | 不适用 | |
| AWS | 南美 / 亚太 / 中东 | 1.2 | 不适用 | |
| Anthropic Claude 3 ↗ | AWS | 北美 | 52 | 258 |
| AWS | 欧盟 / 英国 | 44 | 218 | |
| AWS | 南美 / 亚太 / 中东 | 35 | 177 | |
| Anthropic Claude 3 Haiku ↗ | AWS | 北美 | 4.3 | 21.5 |
| AWS | 欧盟 / 英国 | 3.6 | 18.2 | |
| AWS | 南美 / 亚太 / 中东 | 3.0 | 14.8 | |
| Anthropic Claude 3.5 Haiku ↗ | AWS | 北美 | 12 | 62 |
| AWS | 欧盟 / 英国 | 10 | 52 | |
| AWS | 南美 / 亚太 / 中东 | 9 | 43 | |
| Anthropic Claude 4.5 Haiku ↗ | AWS | 北美 | 17.3 | 86.4 |
| AWS | 欧盟 / 英国 | 14.6 | 73.1 | |
| AWS | 南美 / 亚太 / 中东 | 11.9 | 59.4 | |
| Anthropic Claude 3.5 Sonnet ↗ | AWS | 北美 | 52 | 258 |
| AWS | 欧盟 / 英国 | 44 | 218 | |
| AWS | 南美 / 亚太 / 中东 | 35 | 177 | |
| Anthropic Claude 3.5 Sonnet v2 ↗ | AWS | 北美 | 46 | 232 |
| AWS | 欧盟 / 英国 | 39 | 196 | |
| AWS | 南美 / 亚太 / 中东 | 32 | 159 | |
| Anthropic Claude 4 Sonnet ↗ | AWS | 北美 | 46.4 | 231.8 |
| AWS | 欧盟 / 英国 | 39.2 | 196.2 | |
| AWS | 南美 / 亚太 / 中东 | 31.9 | 159.4 | |
| Anthropic Claude 4.5 Sonnet ↗ | AWS | 北美 | 51.8 | 259.1 |
| AWS | 欧盟 / 英国 | 43.8 | 219.2 | |
| AWS | 南美 / 亚太 / 中东 | 35.6 | 178.1 | |
| Anthropic Claude 4.6 Sonnet ↗ | AWS | 北美 | 54.5 | 272.7 |
| AWS | 欧盟 / 英国 | 46.2 | 230.8 | |
| AWS | 南美 / 亚太 / 中东 | 37.5 | 187.5 | |
| Anthropic Claude 4 Opus ↗ | AWS | 北美 | 232 | 1159 |
| AWS | 欧盟 / 英国 | 196 | 981 | |
| AWS | 南美 / 亚太 / 中东 | 159 | 797 | |
| Anthropic Claude 4.1 Opus ↗ | AWS | 北美 | 259 | 1295 |
| AWS | 欧盟 / 英国 | 219 | 1096 | |
| AWS | 南美 / 亚太 / 中东 | 178 | 891 | |
| Anthropic Claude 4.5 Opus ↗ | AWS | 北美 | 90.9 | 454.5 |
| AWS | 欧盟 / 英国 | 76.9 | 384.6 | |
| AWS | 南美 / 亚太 / 中东 | 62.5 | 312.5 | |
| Anthropic Claude 4.6 Opus ↗ | AWS | 北美 | 90.9 | 454.5 |
| AWS | 欧盟 / 英国 | 76.9 | 384.6 | |
| AWS | 南美 / 亚太 / 中东 | 62.5 | 312.5 | |
| Anthropic Claude 4.7 Opus ↗ | AWS | 北美 | 90.9 | 454.5 |
| AWS | 欧盟 / 英国 | 76.9 | 384.6 | |
| AWS | 南美 / 亚太 / 中东 | 62.5 | 312.5 | |
| Anthropic Claude 4.8 Opus ↗ | AWS | 北美 | 90.9 | 454.5 |
| AWS | 欧盟 / 英国 | 76.9 | 384.6 | |
| AWS | 南美 / 亚太 / 中东 | 62.5 | 312.5 | |
| Mistral Small 24B ↗ | AWS | 北美 | 158 | 525 |
| AWS | 欧盟 / 英国 | 133 | 444 | |
| AWS | 南美 / 亚太 / 中东 | 108 | 361 | |
| Mistral Small 24B Instruct ↗ | AWS | 北美 | 157.5 | 525 |
| AWS | 欧盟 / 英国 | 133.3 | 444.2 | |
| AWS | 南美 / 亚太 / 中东 | 108.3 | 360.9 | |
| Llama 3.1_8B ↗ | AWS | 北美 | 158 | 525 |
| AWS | 欧盟 / 英国 | 133 | 444 | |
| AWS | 南美 / 亚太 / 中东 | 108 | 361 | |
| Llama 3.3_70B ↗ | AWS | 北美 | 158 | 525 |
| AWS | 欧盟 / 英国 | 133 | 444 | |
| AWS | 南美 / 亚太 / 中东 | 108 | 361 | |
| Llama 3.3 70B Instruct ↗ | AWS | 北美 | 157.5 | 525 |
| AWS | 欧盟 / 英国 | 133.3 | 444.2 | |
| AWS | 南美 / 亚太 / 中东 | 108.3 |