narev-lookup-llm-pricing

Compare original and translation side by side

🇺🇸

Original

English

🇨🇳

Translation

Chinese

Look up LLM pricing

查询LLM定价

This skill is the in-repo API reference for the Narev Cloud Pricing endpoints (same behavior as

/platform/api-reference/endpoint/pricing/...

, tightened for agents). Use it for contracts and workflows; for patterns that write the catalog into the repo, see

update-llm-pricing

Two endpoints, both under

https://www.narev.ai

Endpoint	Method	Purpose
`/api/models/pricing`	`GET`	List the catalog. Filter by `model_id` , `search` , `provider` , `subprovider` . Paginated.
`/api/models/pricing/calculate`	`POST`	Compute the USD cost of one call given `modelId` , `provider` , and a `usage` object.

Token rates are USD per token (not per 1K, not per 1M).

本技能是Narev云定价端点的仓库内API参考文档（与

/platform/api-reference/endpoint/pricing/...

的行为一致，专为Agent优化）。适用于合同与工作流场景；若需将定价目录写入仓库的模式，请查看

update-llm-pricing

。

两个端点均位于

https://www.narev.ai

域名下。

端点地址	请求方法	用途
`/api/models/pricing`	`GET`	列出定价目录。可通过 `model_id` 、 `search` 、 `provider` 、 `subprovider` 进行筛选，支持分页。
`/api/models/pricing/calculate`	`POST`	根据 `modelId` 、 `provider` 和 `usage` 对象计算单次调用的美元成本。

Token费率为每token对应的美元价格（非每千token或每百万token）。

When to use this skill

何时使用本技能

"What does
```
gpt-4o
```
cost on OpenAI right now?"
"Calculate how much this prompt cost me."
"How much do cached input tokens cost on Anthropic for
```
claude-sonnet-4
```
?"
"List the providers that serve
```
llama-3.1-70b
```
and their rates."
Any flow that needs token-to-dollar conversion against a current catalog.

If the user wants to automate fetching the catalog and commit mapped prices (offline or deterministic billing), switch to

update-llm-pricing

— it builds on the

GET

behavior this page defines. If they want prices to resolve at runtime inside their app via the SDK, point them at

add-usage-based-billing

"当前OpenAI上的
```
gpt-4o
```
定价是多少？"
"计算这个提示词调用花费了多少钱。"
"Anthropic的
```
claude-sonnet-4
```
中缓存输入token的成本是多少？"
"列出提供
```
llama-3.1-70b
```
的服务商及其费率。"
任何需要基于当前目录进行token转美元换算的流程。

如果用户想要自动化获取定价目录并提交映射后的价格（用于离线或确定性计费），请切换至

update-llm-pricing

——它基于本页定义的

GET

行为构建。如果用户希望在应用运行时通过SDK解析价格，请引导他们使用

add-usage-based-billing

。

Inputs you need

所需输入

The pricing endpoints are public. No API key, bearer token, or authentication header is required — call them directly.

For listing: optional

model_id

search

provider

subprovider

sort_by

(

model_id

provider

subprovider

order

(

asc

desc

page

limit

(max

, default

For calculation:
```
modelId
```
,
```
provider
```
, and
```
usage
```
with
```
promptTokens
```
,
```
completionTokens
```
,
```
cacheReadTokens
```
,
```
cacheWriteTokens
```
,
```
reasoningTokens
```
(all required integers — pass
```
0
```
if unused).
```
subprovider
```
is required when one model is served by multiple providers (
```
bedrock
```
,
```
openrouter
```
,
```
together
```
, etc.).
```
webSearchCount
```
and
```
isByok
```
are optional.

定价端点为公开接口，无需API密钥、Bearer令牌或认证头——可直接调用。

列表查询：可选参数包括
```
model_id
```
、
```
search
```
、
```
provider
```
、
```
subprovider
```
、
```
sort_by
```
（可选值：
```
model_id
```
|
```
provider
```
|
```
subprovider
```
）、
```
order
```
（可选值：
```
asc
```
|
```
desc
```
）、
```
page
```
、
```
limit
```
（最大值
```
1000
```
，默认值
```
100
```
）。
成本计算：必填参数为
```
modelId
```
、
```
provider
```
，以及包含
```
promptTokens
```
、
```
completionTokens
```
、
```
cacheReadTokens
```
、
```
cacheWriteTokens
```
、
```
reasoningTokens
```
的
```
usage
```
对象（所有参数均为必填整数——若未使用则传入
```
0
```
）。当同一模型由多个服务商（如
```
bedrock
```
、
```
openrouter
```
、
```
together
```
等）提供时，
```
subprovider
```
为必填参数。
```
webSearchCount
```
和
```
isByok
```
为可选参数。

Workflow: list pricing

工作流：查询定价列表

Confirm the model and (if needed) the provider. For "OpenAI's GPT-4o", filter by
```
model_id=gpt-4o
```
and
```
provider=openai
```
.

Call the listing endpoint:

bash

curl -G 'https://www.narev.ai/api/models/pricing' \
  --data-urlencode 'model_id=gpt-4o' \
  --data-urlencode 'provider=openai'

The response is
```
{ data: ModelPricingEntry[], meta: { page, limit, total, total_pages } }
```
. Each entry has
```
model_id
```
,
```
provider
```
,
```
subprovider
```
, and a
```
pricing
```
object. The fields you most likely care about:
- ```
price_prompt
```
  — USD per input token.
- ```
price_completion
```
  — USD per output token.
- ```
price_input_cache_read
```
  ,
```
price_input_cache_write
```
  — USD per cached input token.
- ```
price_internal_reasoning
```
  — USD per reasoning output token.
- ```
pricing_request
```
  — flat USD per request.
- ```
price_web_search
```
  — USD per web-search invocation.
- ```
pricing_discount
```
  — fractional discount (
```
0
```
  –
```
1
```
  ) applied across all rates.
- ```
price_image
```
  ,
```
price_image_output
```
  ,
```
price_audio
```
  ,
```
price_audio_output
```
  ,
```
price_input_audio_cache
```
  — USD per unit, when applicable.
Surface the requested fields. Multiply by
```
1_000_000
```
if the user expects "USD per million tokens".
If
```
meta.total_pages > 1
```
, page through with
```
page=2
```
,
```
page=3
```
, …. Filter rather than paginating the whole catalog whenever possible.

确认模型及（如需）服务商。例如查询“OpenAI的GPT-4o”，需通过
```
model_id=gpt-4o
```
和
```
provider=openai
```
进行筛选。

调用列表查询端点：

bash

curl -G 'https://www.narev.ai/api/models/pricing' \
  --data-urlencode 'model_id=gpt-4o' \
  --data-urlencode 'provider=openai'

响应格式为
```
{ data: ModelPricingEntry[], meta: { page, limit, total, total_pages } }
```
。每个条目包含
```
model_id
```
、
```
provider
```
、
```
subprovider
```
及
```
pricing
```
对象。最常用的字段包括：
- ```
price_prompt
```
  —— 输入token对应的美元价格。
- ```
price_completion
```
  —— 输出token对应的美元价格。
- ```
price_input_cache_read
```
  、
```
price_input_cache_write
```
  —— 缓存输入token对应的美元价格。
- ```
price_internal_reasoning
```
  —— 推理输出token对应的美元价格。
- ```
pricing_request
```
  —— 每次请求的固定美元费用。
- ```
price_web_search
```
  —— 每次网页搜索调用的美元费用。
- ```
pricing_discount
```
  —— 适用于所有费率的折扣比例（取值范围
```
0
```
  –
```
1
```
  ）。
- ```
price_image
```
  、
```
price_image_output
```
  、
```
price_audio
```
  、
```
price_audio_output
```
  、
```
price_input_audio_cache
```
  —— 对应单位的美元价格（适用时显示）。
展示用户所需的字段。如果用户期望“每百万token的美元价格”，需将数值乘以
```
1_000_000
```
。
若
```
meta.total_pages > 1
```
，需通过
```
page=2
```
、
```
page=3
```
……进行分页查询。尽可能优先筛选而非遍历整个目录。

Workflow: calculate the cost of a call

工作流：计算单次调用成本

Collect token usage. Required integers:
```
promptTokens
```
,
```
completionTokens
```
,
```
cacheReadTokens
```
,
```
cacheWriteTokens
```
,
```
reasoningTokens
```
. If a category does not apply, pass
```
0
```
.

POST the request:

bash

curl -X POST 'https://www.narev.ai/api/models/pricing/calculate' \
  -H 'Content-Type: application/json' \
  -d '{
    "modelId": "gpt-4o",
    "provider": "openai",
    "subprovider": "OpenAI",
    "usage": {
      "promptTokens": 1000,
      "completionTokens": 500,
      "cacheReadTokens": 0,
      "cacheWriteTokens": 0,
      "reasoningTokens": 0
    }
  }'

The 200 response contains:
- ```
pricing
```
  — the rates Narev applied (
```
input
```
  ,
```
output
```
  ,
```
request
```
  ,
```
inputCacheRead
```
  ,
```
inputCacheWrite
```
  ,
```
internalReasoning
```
  ,
```
webSearch
```
  ).
- ```
costBreakdown.total
```
  — the final USD total.
- ```
usage
```
  — echoed back so the caller can verify what was billed.
Report
```
costBreakdown.total
```
to the user. If they ask "where does that number come from?", show
```
pricing
```
and explain that each token category was multiplied by its rate and summed.

收集token使用数据。必填整数参数：
```
promptTokens
```
、
```
completionTokens
```
、
```
cacheReadTokens
```
、
```
cacheWriteTokens
```
、
```
reasoningTokens
```
。若某类别不适用，传入
```
0
```
。

发送POST请求：

bash

curl -X POST 'https://www.narev.ai/api/models/pricing/calculate' \
  -H 'Content-Type: application/json' \
  -d '{
    "modelId": "gpt-4o",
    "provider": "openai",
    "subprovider": "OpenAI",
    "usage": {
      "promptTokens": 1000,
      "completionTokens": 500,
      "cacheReadTokens": 0,
      "cacheWriteTokens": 0,
      "reasoningTokens": 0
    }
  }'

200响应包含以下内容：
- ```
pricing
```
  —— Narev应用的费率（包括
```
input
```
  、
```
output
```
  、
```
request
```
  、
```
inputCacheRead
```
  、
```
inputCacheWrite
```
  、
```
internalReasoning
```
  、
```
webSearch
```
  ）。
- ```
costBreakdown.total
```
  —— 最终的美元总成本。
- ```
usage
```
  —— 回传的使用数据，便于调用方验证计费依据。
向用户展示
```
costBreakdown.total
```
。如果用户询问“这个数值是如何计算的？”，请展示
```
pricing
```
并说明每个token类别均乘以对应费率后求和。

Constraints and edge cases

约束条件与边缘情况

402 Payment Required — model is enterprise-only. The error response has
```
error
```
set and
```
pricing: null
```
. Tell the user the model is not in the public catalog and point them at Narev for enterprise access.
404 Not Found — no public pricing for that
```
modelId
```
+
```
provider
```
(+
```
subprovider
```
). Re-check IDs against the listing endpoint or drop
```
subprovider
```
.
400 Bad Request —
```
usage
```
failed validation. Most common cause: a missing required integer field. Pass
```
0
```
, not
```
null
```
or absent.
Rates are USD per token. Do not divide or multiply on the way in. Convert only when displaying.
subprovider
matters. Models like
```
llama-3.1-70b
```
are hosted by Bedrock, OpenRouter, Together, etc. at different prices. If the user does not know the subprovider, list first to see options, then calculate.
Do not hardcode rates. Prices change. If the user wants stable values they can commit and review, switch to
```
update-llm-pricing
```
.

402 Payment Required —— 模型为企业专属。错误响应中
```
error
```
字段会被设置，
```
pricing
```
为
```
null
```
。告知用户该模型不在公开目录中，并引导他们联系Narev获取企业权限。
404 Not Found —— 该
```
modelId
```
+
```
provider
```
（+
```
subprovider
```
）组合无公开定价。请通过列表查询端点重新核对ID，或移除
```
subprovider
```
参数重试。
400 Bad Request ——
```
usage
```
参数验证失败。最常见原因是缺少必填整数字段。请传入
```
0
```
，而非
```
null
```
或留空。
费率为每token美元价格。调用时请勿进行除法或乘法运算，仅在展示时进行转换。
subprovider
至关重要。像
```
llama-3.1-70b
```
这样的模型由Bedrock、OpenRouter、Together等不同服务商托管，定价各不相同。如果用户不知道subprovider，先查询列表查看可选选项，再进行成本计算。
请勿硬编码费率。价格会发生变动。如果用户需要可提交并审核的稳定数值，请切换至
```
update-llm-pricing
```
。

Reference

参考链接

API overview:
```
/platform/api-reference/introduction
```

List endpoint:

/platform/api-reference/endpoint/pricing/list-model-pricing

Calculate endpoint:

/platform/api-reference/endpoint/pricing/calculate-cost-for-a-model-call

API概览：
```
/platform/api-reference/introduction
```

列表查询端点：

/platform/api-reference/endpoint/pricing/list-model-pricing

成本计算端点：

/platform/api-reference/endpoint/pricing/calculate-cost-for-a-model-call