Skip to content
← Models
Open weightsSnowflake/10 trustVerified
Snowflake Arctic Embed 2.0 logo

Snowflake Arctic Embed 2.0

Apache 2.0 multilingual embedding that keeps English quality high.

Visit official page →

About

Arctic Embed 2.0 shipped in December 2024 in medium (305M) and large (568M) variants. Snowflake designed it to avoid the usual multilingual tradeoff: it tops English retrieval while also covering roughly 50 to 70 languages, with 8K context and Matryoshka truncation. Apache 2.0 licensed with ONNX exports, it is a common pick for teams that want strong open-weight retrieval without LLM-scale hardware.

Specs

Context window

8,192 tokens

Pricing

Free (self-hosted)

Free quota

Free via self-hosting

API access

Self-host (Hugging Face, Ollama, TEI)

Weights

Open weights

Released

December 7, 2024

Benchmarks

Team-reported: MTEB Retrieval nDCG@10 55.98 (large variant)

Provider-reported unless a source says otherwise. Benchmarks are marketing-adjacent by nature — treat them as directional.

Use cases

multilingual-retrievallocal-embeddingsrag

Trust score — 8/10

/10 trustVerified

Scored 8/10 (verified tier, metrics-v1). Evidence: Release 2024-12-07 verified via MarkTechPost; Apache 2.0, 8192 tokens, multilingual, MTEB Retrieval 55.98 (L) verified via Snowflake's own engineering blog and a GGUF model card.; Team-reported claims about no-English-degradation come from Snowflake's blog; noted as vendor claims.; Medium variant is GTE-multilingual based (305M); covered in description only..

Sources