Pricing & balance

All Writer bills AI usage in CNY (yuan) as a single account balance. Every AI feature draws from that balance.

Units

• CNY (yuan) — what you see in the product and what is charged.
• Tokens — model input/output units; input tokens can be split into uncached input and cached input.

Charges are computed per million tokens for uncached input, cached input, and output separately:

Cost (CNY)
= uncached_input_tokens × input_price_per_million
+ cached_input_tokens × cached_input_price_per_million
+ output_tokens × output_price_per_million

Cached input means prompt tokens reported by the model provider as cache hits. Cache-hit input is billed with its own cached-input price, while the remaining input uses the normal input price. Output is often more expensive per token than input, so longer answers cost more even with the same prompt.

The current supported official Chat model is deepseek-v4-pro: cached input is ¥0.025 per million tokens, uncached input is ¥3 per million tokens, and output is ¥6 per million tokens. Custom Chat models use your own upstream provider and API key; upstream model charges are handled by that provider, while any All Writer platform abilities used in the same task still consume account balance.

Web search and URL-to-Markdown are search-style per-call abilities. Web search costs ¥0.01 per call, and URL-to-Markdown costs ¥0.01 per call; knowledge-base search and AI Word export still consume balance based on their actual usage.

Packages

Available balance packages are shown in Settings → Account. A package normally shows the amount you pay, the balance credited to your account, and any bonus credit when applicable.

Self-service online checkout is not available at the moment. Top-ups are handled manually through support WeChat (see Community, which points to the Chinese page for the QR code). This page and the product UI will be updated if self-service payment returns.

Where to see usage

Open Settings → Account to view balance and recent usage (including search, URL-to-Markdown, and knowledge-base search where applicable).

Search, URLs, and Word export

• @web: web search costs ¥0.01 per call; cached repeats usually avoid repeated charge.
• URL-to-Markdown: fetching a page body as Markdown costs ¥0.01 per call.
• @base: knowledge-base search usually costs less than long generation, but still consumes balance.
• AI Word export: planning, polishing/rewriting, summaries, and other model calls consume balance based on actual usage.

Custom Chat models

Custom Chat models use the Base URL and API key you configure in Settings → AI Features. The upstream provider handles the model bill. If an Agent task also uses @web, @base, URL-to-Markdown, or AI Word export, those All Writer platform abilities still consume your All Writer balance.

Need more detail?

• AI features — what the Agent can do
• Getting started — first steps

For the full Chinese billing page (tables and examples), see 计费与套餐.