Claude API New Features at a Glance: Long Outputs, Workbench Evaluations, and a Usage Dashboard

Claude API’s latest wave of updates is more geared toward “everyday developer efficiency.” From longer outputs in Claude Sonnet 3.5, to prompt generation and evaluation in the Claude Console Workbench, to clearer usage and cost tracking, these changes can directly reduce the time spent repeatedly tuning parameters and reconciling bills.

Claude API Long Outputs: Sonnet 3.5 Limit Doubled

In the Claude API, the maximum output tokens for Claude Sonnet 3.5 has been increased from 4096 to 8192, making long-form summaries, code generation, and multi-turn reasoning less likely to get “cut off mid-sentence.”

How to enable it is also straightforward: when calling the Claude API, add the request header "anthropic-beta": "max-tokens-3-5-sonnet-2024-07-15", and use it together with your existing max_tokens setting.

Claude Console Workbench: The Prompt Generator Is More Like a “Drafting Assistant”

If you often prototype prompts before using the Claude API, the Workbench’s new “Prompt Generator” is a real time-saver: just describe the task in one sentence (for example, “route inbound customer support tickets and assign priorities”), and Claude will produce a more structurally complete prompt draft.

It doesn’t replace your domain rules, but it can lay down a first draft of the “role, constraints, output format, and examples,” reducing the time spent starting from a blank document.

Evaluation Mode: Compare Prompts Side by Side—Less Guesswork, More Results

The Workbench’s “Evaluation Mode” lets you view the outputs of two or more prompts side by side and score Claude’s output on a 5-point scale. This is especially useful for prompt iteration, A/B testing, or when turning rules into production prompts for the Claude API.

Especially when you’re choosing between “shorter and cheaper in tokens” and “more stable with fewer false positives,” this kind of visual comparison helps you set direction faster.

Usage and Cost Dashboard: Claude API Costs Are Finally Easy to Reconcile

The developer console has added “Usage” and “Cost” tabs, allowing you to track Claude API usage and billing by dollar amount, token count, and API key. For multi-project, multi-environment (staging/production) setups, it becomes much more straightforward to pinpoint which key is spending faster.

In addition, the official documentation has been expanded with more comprehensive release notes and learning resources (courses and Cookbook guides), making it easier to take your Claude API usage from “it runs” to “reusable and maintainable.”