nomic-embed-text:v2-moe
Upstream reference: Nomic Embed Text v2 MoE
Publisher: Nomic AI
Nomic describes a multilingual mixture-of-experts text encoder for retrieval, using task prefixes and adaptable vector dimensions.
Publisher-described context, separate from Sigma5C observations and recorded catalog statistics.
Published upstream specifications
- Total parameters
- 475 million Source
- Active parameters
- 305 million Source
- Published sequence length
- 512 tokens Source
- Vector dimensions
- 768, reducible to 256 with Matryoshka truncation Source
- Upstream license label
- Apache-2.0 Source
Publisher-described uses
- Multilingual document and query retrieval. Source
Reference limitations
- The v1 encoder’s 8192-token specification must not be substituted for the v2 MoE card’s 512-token maximum. Source
A recorded tag and an upstream card association do not verify the deployed checkpoint, quantization, configured context or current hosted availability.
Publisher sources
- Nomic Embed Text v2 MoE official documentation — Nomic AI. Checked .
Links
Recorded references do not establish downloadable weights, licensing or current availability.
2048
context
apache-2.0
license
2024-02-10
released
Where nomic-embed-text:v2-moe comes from
Company / lab
Nomic AI
See every model and family from this maker.
Model family
nomic-embed-text
Other versions and sizes in this line.
embedding