Voyage voyage-3-large
Voyage AI
Capabilities
Strengths
- Top MTEB retrieval at release (>74 average)
- Matryoshka — 256/512/1024/2048 dim options
- Long 32k input window
Weaknesses
- Closed weights
- Smaller provider — fewer SDK integrations
Pricing
Input / 1M tokens
$0.18
Output
—
Hosting
Hosted only
Embedding specs
Output dimension
1024
Max input
32,000 tokens
Matryoshka
Supported
Matryoshka representation learning: dimension can be truncated to a smaller size at index time without retraining, trading retrieval quality for index size and speed.
Transparency
Open weights
0.0 / 10
Open training data
1.0 / 10
Open methodology
2.0 / 10
Licence openness
1.0 / 10
Provider disclosure
4.0 / 10
FMTI company score
N/A
Closed weights; some methodology disclosed.
Sustainability
Inference energy
N/A
Training footprint
N/A
Provider infrastructure
3.0 / 10
No published sustainability data.
MTEB quality
Grounded in the MTEB (Massive Text Embedding Benchmark) Overall average published by the model authors. Bearing collapses MTEB's retrieval, STS, classification, and clustering categories into a single quality signal because they correlate strongly for embedding models. Methodology.