Vercel ai@7.0.129 Adds 'max' Reasoning Level and Deprecation Warnings

What changed?

Vercel's ai SDK version 7.0.129 introduces a portable 'max' reasoning level for language model calls. This new level is mapped appropriately to provider-native options or handled with fallbacks based on available capabilities and budgets. Deprecation warnings are now emitted with stable codes and are deduplicated, making it easier to manage code maintenance. The release also includes several minor fixes and upgrades, such as improved error publishing, provider option preservation, WebKit network error recognition, and an upgrade to TypeScript 6.

This update modifies the TypeScript types for reasoning levels. Any exhaustive switch or Record logic on reasoning values must be updated to handle the new 'max' case. Compatibility notes for Amazon Bedrock and Fireworks provider mappings are provided, with compatibility warnings emitted as needed.

Why does it matter to an everyday developer?

Developers now have a standardized way to request the highest available reasoning effort across multiple AI providers without having to manage provider-specific configuration. The stable and deduplicated deprecation warnings improve code maintainability by making deprecation tracking more systematic. However, those using exhaustive TypeScript logic to process reasoning levels must add explicit support for the new 'max' value to avoid runtime issues or type errors.

What can the developer do now?

If you use provider reasoning levels in Vercel's ai SDK, review your TypeScript code for exhaustive switches or Record mappings over reasoning values. Add handling for the new 'max' value before upgrading to avoid source compatibility issues. You can now utilize the 'max' reasoning level to request top-tier AI capabilities, benefitting from automatic provider compatibility and fallback support. Monitor your logs for new deprecation warning codes for easier future-proofing.