The full upstream README, mirrored here for reference. Install config, tool schemas, adoption signals, and an original overview live on the Llm Prices Cn listing page.
This repository provides a daily-verified pricing dataset of popular Large Language Model (LLM) APIs, tracking both Chinese and global model providers.
This dataset is licensed under the Creative Commons Attribution 4.0 International License (CC-BY-4.0).
Under this license, you are free to:
Attribution Requirements:
You must give appropriate credit and provide a link back to the original source. When using this dataset in your projects, articles, or websites, you must include a clickable hyperlink to LLM Abacus or https://llmabacus.com.
For example:
Pricing data provided by LLM Abacus.
prices.json: The core dataset containing structured model details, metadata, and token prices (both USD and CNY per million tokens).sync_prices.py: A utility script to synchronize and rebuild prices.json from the source application configurations.CONTRIBUTING.md: Guidelines on how to report price discrepancies or request new model tracking.prices.json)The generated prices.json file uses the following schema:
last_updated: YYYY-MM-DD date when the pricing was last updated.usd_to_cny_rate: The exchange rate used for conversions.pricing_unit: Pricing metrics unit (default is per_million_tokens).models: Array of model objects:
id: Unique model identifier (e.g., gpt-5-5).name: Human-readable display name.vendor_id / vendor_name: Vendor classification.country: Region of origin (e.g., US, CN).billing_currency: Original currency of vendor billing (USD or CNY).input_price_usd_per_m / output_price_usd_per_m: Price per million tokens in USD.input_price_cny_per_m / output_price_cny_per_m: Price per million tokens in CNY.cached_input_price_usd_per_m / cached_input_price_cny_per_m: Optional. Cache hit pricing.context_window: Maximum input context length.max_output: Maximum generation output length.modality: Supported modalities (text, vision, etc.).tags: Classification tags (e.g., 旗舰, 推理, 性价比).knowledge_cutoff: Model knowledge cutoff date.quality_score: Benchmarked quality index.To update prices.json with the latest upstream data, run:
本仓库提供每日核价的国内外主流大语言模型 (LLM) API 价格数据集。
本数据集采用 知识共享署名 4.0 国际许可协议 (CC-BY-4.0) 进行许可。
您可以自由地:
署名与回链要求:
您必须给出适当的署名,并提供指向原始主站的链接。在您的项目、文章或网页中引用本数据时,必须保留指向 LLM Abacus (https://llmabacus.com) 的超链接。
示例:
价格数据来源于 LLM Abacus。
prices.json:包含核心模型结构、元数据及百万 token 价格(换算为 USD 和 CNY)的 JSON 快照。sync_prices.py:从主站配置同步与生成最新 prices.json 的 Python 脚本。CONTRIBUTING.md:贡献指南,指导如何纠错或申请收录新模型。prices.json)last_updated:最后更新日期 (YYYY-MM-DD)。usd_to_cny_rate:核价所采用的美元兑人民币汇率。pricing_unit:计费单位(固定为 per_million_tokens 即每百万 token)。models:模型数组:
id:模型唯一标识符 (例如 gpt-5-5)。name:模型显示名称。vendor_id / vendor_name:厂商标识与名称。country:厂商归属国家 (如 US, CN)。billing_currency:官方计费结算货币 (USD 或 CNY)。input_price_usd_per_m / output_price_usd_per_m:每百万 input/output token 折算美元价。input_price_cny_per_m / output_price_cny_per_m:每百万 input/output token 折算人民币价。cached_input_price_usd_per_m / cached_input_price_cny_per_m:提示词缓存命中时的单价(如有)。context_window:上下文窗口大小。max_output:最大单次输出限制。modality:支持模态 (text, vision 等)。tags:模型特征标签(如旗舰、推理、性价比)。knowledge_cutoff:知识截止时间。quality_score:基准评测质量分。如需从主站代码库更新并重新构建 prices.json,请执行:
This dataset is also exposed as a remote MCP server so AI agents can query live pricing and estimate token costs directly:
https://www.llmabacus.com/api/mcp/mcpquery_model_price(model), estimate_cost(text_or_tokens, model)You can also run this dataset as a local stdio MCP server — it reads the bundled prices.json, needs no network, and works offline:
Tools: list_llm_prices(vendor?, currency?), estimate_cost(model_id, input_tokens, output_tokens, currency?).