我比较了五种跟踪 LLM API 成本的方法（只有一种能实时工作）

For users on pay-per-token plans, unexpected bills can be a common issue. The main options for tracking costs include: 1. Provider Dashboards (like OpenAI and Anthropic) that update every few hours but lack per-session breakdowns; they are free but can lead to overspending by the time users check. 2. LiteLLM Proxy, which tracks costs across multiple providers, requires self-hosting, is free and open-source, but involves setup overhead and does not provide real-time visibility. 3. Helicone offers great logging and analytics but requires routing traffic through its proxy, has a free tier, and is dashboard-based, lacking real-time feedback. Each option has its pros and cons, and users need to choose based on their specific needs.