Skip to content

feat: extended token usage#10975

Merged
lgrammel merged 44 commits intomainfrom
lg/XY180bQG
Dec 9, 2025
Merged

feat: extended token usage#10975
lgrammel merged 44 commits intomainfrom
lg/XY180bQG

Conversation

@lgrammel
Copy link
Copy Markdown
Collaborator

@lgrammel lgrammel commented Dec 8, 2025

Background

Providers have added extended usage information over the past year, and current usage calculations are inconsistent (see #9921 ).

Summary

  • restructure LanguageModelV3Usage in specification to include caching, reasoning, and raw usage.
  • expose additional inputTokenDetails and outputTokenDetails in LanguageModelUsage
  • deprecate reasoningTokens and cachedInputTokens in LanguageModelUsage

Manual Verification

  • run examples/ai-core/src/generate-text/openai.ts
  • run examples/ai-core/src/stream-text/openai.ts

Tasks

  • spec
  • ai return
  • provider implementation
  • model v2 adapter generate
  • model v2 adapter stream
  • docs
  • migration guide
  • changeset

Future Work

  • add codemod for rewriting access to deprecated LanguageModelUsage properties
  • ensure that raw anthropic usage information is correctly forwarded
  • language model v2 mapping: map stream parts as needed

Related Issues

Resolves #9921

@vercel-ai-sdk vercel-ai-sdk bot added the ai/core core functions like generateText, streamText, etc. Provider utils, and provider spec. label Dec 8, 2025
@lgrammel lgrammel changed the title 1 feat: extended token usage Dec 8, 2025
@lgrammel lgrammel marked this pull request as ready for review December 9, 2025 13:23
@lgrammel lgrammel merged commit 3bd2689 into main Dec 9, 2025
18 checks passed
@lgrammel lgrammel deleted the lg/XY180bQG branch December 9, 2025 15:03
nicoalbanese added a commit that referenced this pull request Dec 9, 2025
#10975 broke submodule action. This should fix.
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

ai/core core functions like generateText, streamText, etc. Provider utils, and provider spec.

Projects

None yet

Development

Successfully merging this pull request may close these issues.

V3 Spec Proposal: Token Usage Normalization for Vercel AI SDK

3 participants