Nomic Embed Text v2
First mixture-of-experts embedding model, multilingual and fully open.
About
Nomic's second-generation embedding model and the first general-purpose MoE embedder: 475M total parameters with only 305M active per token, trained on 1.6B pairs across roughly 100 languages.
Matryoshka embeddings let you truncate the 768-dim output to 256 dims with modest quality loss. Weights, training data, and code are all Apache 2.0 — the full reproducibility package.
Specs
Context window
Pricing
Free quota
API access
Weights
Released
Benchmarks
State of the art on MIRACL for its size; outperforms v1.5 on BEIR and MIRACL per Nomic.
Provider-reported unless a source says otherwise. Benchmarks are marketing-adjacent by nature — treat them as directional.
Use cases
Trust score — 8/10
Scored 8/10 (verified tier, metrics-v1). Evidence: Strong adoption since the February 2025 release; first MoE embedding model drew wide attention; Nomic maintains the model with GGUF and ONNX-adjacent community ports; Weights, training data, and code all Apache 2.0. Manual review: 475M/305M parameter split and Matryoshka truncation verified on the model card and paper. License: Apache 2.0.
How the trust score is built
Full methodology →- Adoption · stars/downloads, log-scaled vs peers
- 7.0/10
- Maintenance · recency, releases, issue load
- 7.0/10
- Openness · license + docs
- 10.0/10
- Manual review · sanity read + red-flag check
- 8.0/10
- · Strong adoption since the February 2025 release; first MoE embedding model drew wide attention
- · Nomic maintains the model with GGUF and ONNX-adjacent community ports
- · Weights, training data, and code all Apache 2.0
- · manual review: 475M/305M parameter split and Matryoshka truncation verified on the model card and paper