Introduction to new Claude API features: Models API, extended output, and the cost dashboard

This round of Claude API updates is geared more toward “everyday developer usefulness.” The core is making model discovery, long outputs, and usage billing more controllable. This article breaks down the Models API, the increased output limit, and the console’s usage and cost dashboards, so you can plug them directly into your existing calling workflow.

Models API: Check available models before making a request

In the Claude API, the value of the Models API is straightforward: you can query the currently available models and verify that the model ID you plan to use is correct. For multi-environment deployments, this reduces production issues like “model unavailable” or “wrong ID,” shifting validation earlier into the release pipeline.

If you have multiple API keys or multiple projects, it’s recommended to fetch the list once during initialization via the Models API and validate it against an allowlist. This way, before your Claude API request enters the main logic, you can confirm the model is available, and your logs will be easier to troubleshoot.

Extended output: Finish long content in one go

Claude API provides extended output for Claude Sonnet 3.5, increasing the maximum output tokens from 4096 to 8192. You enable it by adding a specific request header (anthropic-beta). It’s well-suited to scenarios where “getting cut off midway hurts,” such as long reports, long code generation, or bulk整理 meeting minutes.

In practice, it’s recommended to adjust two things at the same time: first, make the frontend “generating” indicator a continuously streaming display; second, relax the Claude API timeout and retry strategy a bit to avoid long outputs being interrupted by network jitter.

Usage and cost dashboards: Make billing clear

After the developer console added “Usage” and “Cost” tabs, tracking Claude API costs no longer requires cobbling together internal reports. You can view consumption by USD amount, token count, and API key—useful for team cost allocation and investigating abnormal usage.

If you need to align budgets within a company, it’s recommended to use “by API key” as the default management granularity: whose key, which service, and how much it consumed are immediately clear. With Claude API costs made transparent, it also becomes easier to push optimizations like caching, truncation, and slimming down prompts.

More complete release notes: No more guessing updates

Claude API documentation has added more systematic release notes covering changes to the API, the Claude console, and application-side behavior. For development teams, this is much friendlier than “suddenly discovering behavior changed”: you can evaluate the impact in advance and decide whether to upgrade the SDK or adjust parameters.

It’s recommended to incorporate release notes into routine checks: before each iteration, quickly scan the Claude API updates—especially model IDs, output limits, and the console’s billing definitions, which affect stability and cost.