Skip to content

feat(voyage): add voyage-context-4, voyage-4 family, voyage-4-nano and voyage-multimodal-3.5 - #8

Closed
fzowl wants to merge 3 commits into
mainfrom
feat/voyage-context-4
Closed

feat(voyage): add voyage-context-4, voyage-4 family, voyage-4-nano and voyage-multimodal-3.5#8
fzowl wants to merge 3 commits into
mainfrom
feat/voyage-context-4

Conversation

@fzowl

@fzowl fzowl commented Jul 27, 2026

Copy link
Copy Markdown
Owner

What

Adds the latest Voyage AI models and their pricing to the model cost map, and makes the contextual embeddings path robust to standard list[str] inputs.

New models

Model Mode Price /1M tokens Max tokens
voyage/voyage-4 embedding $0.06 32k
voyage/voyage-4-large embedding $0.12 32k
voyage/voyage-4-lite embedding $0.02 32k
voyage/voyage-4-nano embedding $0.00 (free) 32k
voyage/voyage-context-4 contextual embedding $0.12 120k
voyage/voyage-multimodal-3.5 embedding $0.12 32k

voyage-4-nano is Voyage's free/open-weight model and works with the latest voyageai python package.

Contextual embeddings

The contextualized embeddings API expects inputs as list[list[str]] (each document a list[str] of chunks). transform_embedding_request now normalizes a plain string or a flat list[str] into that nested shape, so callers using the standard embedding input still hit the API correctly. Nested inputs pass through unchanged.

Files changed

  • model_prices_and_context_window.json + litellm/model_prices_and_context_window_backup.json — new model entries
  • litellm/llms/voyage/embedding/transformation_contextual.py — input normalization to list[list[str]]
  • docs/my-website/docs/providers/voyage.md — document new models
  • tests/llm_translation/test_voyage_ai.py, tests/test_litellm/llms/voyage/embedding/test_voyage_contextual_transformation.py — tests

Testing

All voyage tests pass.

fzoll and others added 3 commits July 27, 2026 18:07
- Add voyage-context-4 (contextual, 120K max tokens, $0.12/1M)
- Add voyage-4, voyage-4-large, voyage-4-lite, and free voyage-4-nano
- Normalize contextual embedding inputs so a flat list[str] is sent to
  the API as list[list[str]] (each document a list[str] of chunks)
- Update Voyage provider docs and add contextual transformation tests
- Add voyage-multimodal-3.5 embedding model ($0.12/1M, 32k context)
- Add pricing/registration tests for voyage-4 family, voyage-context-4,
  voyage-4-nano, and voyage-multimodal-3.5
@fzowl fzowl closed this Jul 27, 2026
@fzowl
fzowl deleted the feat/voyage-context-4 branch July 27, 2026 17:29
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

2 participants