About this skill

The problem

LLM vendor pricing, context limits, free quotas, and subscription plans are scattered across vendor docs and consoles, often using different units and table layouts. Manually normalizing per 1k tokens versus per 1M tokens, tiered billing, and cache-hit pricing is error-prone. This skill uses vendor pricing pages and Coding Plan pages as data sources, then generates or updates a fixed-structure comparison document for model selection, procurement review, or internal knowledge-base maintenance.

How it works and limits

The skill decides scope from $ARGUMENTS: no argument or all updates all vendors, while a vendor name updates only that section. The workflow includes:
- Read the existing document: if 大模型厂商价格对比完整总结.md exists, inspect its structure before incremental or full updates; otherwise create it from scratch.
- Fetch pricing data: use WebFetch to extract model name, modality, input/output prices, context limits, free quotas, cache-hit prices, and tiered billing rules.
- Normalize formatting: convert prices to CNY per 1M tokens, preserve token ranges, free-model markers, and plan comparisons.
- Write the document: update the top timestamp, replace the full document, or update only the matching vendor section.
- Summarize the run: report updated vendors, failed fetches, and the saved path.

It depends on vendor pages being fetchable; login walls, anti-bot checks, or layout changes can cause a vendor to fail. It maintains a comparison document rather than monitoring prices continuously or making procurement decisions.

Screenshots

Use Cases

Before model selection, compile Aliyun, Zhipu, DeepSeek, and other vendor pricing into one normalized comparison table for procurement review.
When procurement checks subscription plans, refresh the Coding Plan section while preserving the existing document structure and timestamp for auditability.
While maintaining an internal knowledge base, fetch public vendor pricing pages and generate a dated pricing comparison document for repeated reviews.
When comparing multimodal model costs, extract input, output, context limits, free quotas, and normalize price units for model selection spreadsheets.

Best For

Algorithm engineers selecting LLMs who need multiple vendors' per-token pricing in one format.
Procurement engineers verifying input and output prices, free quotas, and subscription plans.
Developers maintaining technical docs who want public pricing pages converted into dated comparison documents.
Platform engineers evaluating multimodal model costs who need context limits and billing rules normalized.