Sign in
← AI PulseJul 27, 2026

Deep · news · Multi-source brief

Vercel AI SDK Adds Experimental Speech Translation and TogetherAI Usage Reporting

The Vercel AI SDK introduces an experimental speech translation model specification and streaming speech-to-speech translation, alongside enabling token usage reporting for TogetherAI streaming responses.

By Illumora Editorial

Source · Jul 27, 2026, 9:55 PM · On Illumora · Jul 27, 2026, 10:08 PM

Media from the primary source — shown here so you can stay on Illumora.

Synthesized from multiple allowlisted primaries on the same event. Lanes →

Brief drafted by Illumora’s editorial model from the linked primary source. Ops desk reviews flagged pieces. How we write →

Read the source →Vercel AI SDK Changelog — Release @ai-sdk/workflow-harness@1.0.44 · vercel/ai
Save

The Vercel AI SDK has released updates including the addition of an experimental speech translation model and enhanced token usage reporting. The updates, detailed in releases for @ai-sdk/workflow-harness@1.0.44, @ai-sdk/workflow@1.0.38, and @ai-sdk/vue@4.0.38, introduce Experimental_SpeechTranslationModelV4 and experimental_streamTranslate for streaming speech-to-speech translation.

Separately, the @ai-sdk/togetherai@1.0.50 release enables includeUsage for TogetherAI, allowing streaming responses to report token usage.

Key Points

  • The Vercel AI SDK now includes an experimental speech translation model specification, Experimental_SpeechTranslationModelV4.
  • An experimental_streamTranslate function supports streaming speech-to-speech translation.
  • The SDK avoids synthesizing client tool errors for invalid provider-executed tool calls.
  • The @ai-sdk/togetherai@1.0.50 release enables includeUsage for TogetherAI.
  • TogetherAI streaming responses can now report token usage.

Context

According to the Vercel AI SDK changelog, the updates to @ai-sdk/workflow-harness, @ai-sdk/workflow, and @ai-sdk/vue share common patch changes. These changes include the introduction of the experimental speech translation features and a refinement in error handling for tool calls. The 7bd6bdd patch specifically addresses avoiding client tool errors for invalid provider-executed tool calls. The 1e2f324 patch introduces the experimental speech translation model and streaming functionality. The @ai-sdk/togetherai update, identified by patch 5dcbf7e, focuses on enabling token usage reporting for TogetherAI's streaming capabilities.

Why It Matters

These updates offer builders new capabilities for integrating speech translation into applications and provide more transparency in token consumption for TogetherAI users. The experimental nature of the speech translation features suggests an opportunity for early testing and feedback, while improved usage reporting can aid in cost management and performance analysis for applications leveraging TogetherAI.

What To Do

  • Note the Experimental_SpeechTranslationModelV4 specification for potential use in speech-to-speech translation applications.
  • Test the experimental_streamTranslate function for streaming speech translation workflows.
  • Review the updated error handling for provider-executed tool calls to understand its impact on client-side error synthesis.
  • If using TogetherAI, verify that includeUsage is enabled to monitor token consumption in streaming responses.