Model price table

This table is auto-generated from /api/pricing (see the build-time sync script). The latest prices are here.

Field meanings:

• Input ($/1M) = USD per 1M prompt tokens
• Output ($/1M) = USD per 1M completion tokens
• Context = max context window
• Capabilities = supported feature tags

Loading from /api/pricing…

How to use this table

1. Find the model ID you want (e.g. deepseek-chat)
2. Read input / output prices
3. Estimate: (prompt tokens / 1M × input price) + (completion tokens / 1M × output price) = per-call cost
4. Monthly: daily calls × 30 × per-call cost

Want cheaper?

The Chinese frontier models on primerouter are notably cost-effective:

• DeepSeek V3 / Reasoner: reasoning ability comparable to GPT-4o, ~1/10 the price
• Qwen Max / Plus: strong on Chinese language tasks
• GLM-4 Plus: solid all-around at friendly prices
• Kimi (Moonshot): ultra-long context (200K+) at low cost

See per-model input / output prices in the table.

Price changes

Upstream price changes / new models / retirements are reflected here:

• Auto-sync runs nightly (CI-triggered)
• Major changes go to Changelog
• Price increases: at least 7 days advance email notice

Want price history?

We don't maintain historical price tables — upstream prices themselves change. Current price is what's on this page.