E5-Mistral 7B
The 7B embedding that topped MTEB — MIT-licensed, from Microsoft Research's E5 team.
About
E5-Mistral-7B-Instruct (early 2024, Microsoft Research via intfloat) was a landmark: a 7B LLM-based embedding model that topped the MTEB leaderboard at release, proving that large decoder models make excellent embedders. Instruction-tuned for retrieval, classification, and clustering.
MIT licensed — commercial use fully allowed. 197k Hugging Face downloads. The 7B size needs serious GPU for embedding workloads (it's a full LLM), which limited adoption vs. smaller models — but for maximum retrieval quality on capable hardware, the LLM-embedder recipe it pioneered (followed by NV-Embed, GTE-Qwen2, SFR) remains the quality ceiling.
Specs
Context window
Pricing
Free quota
API access
Weights
Released
Benchmarks
Provider-claimed at release (2024): #1 on MTEB — treat as provider-claimed and dated; the MTEB top has since moved on. Independent: widely cited as the LLM-embedder pioneer.
Provider-reported unless a source says otherwise. Benchmarks are marketing-adjacent by nature — treat them as directional.
Use cases
Trust score — 7/10
Scored 7/10 (metrics tier). The LLM-embedder pioneer that topped MTEB, MIT-licensed, with lasting methodological influence. Docked three points: (1) 7B is heavy for embedding workloads — GPU-hungry; (2) 2024-era, the MTEB top has moved on; (3) modest adoption (197k downloads) vs. smaller alternatives. For quality-at-any-compute-cost retrieval.
How the trust score is built
Full methodology →- Adoption · stars/downloads, log-scaled vs peers
- 7.3/10
- Maintenance · recency, releases, issue load
- 4.0/10
- Openness · license + docs
- 9.1/10
- Manual review · sanity read + red-flag check
- 7.0/10
- · HF downloads 197,122 (peer-normalized)
- · released 1002 days ago
- · manual review: LLM-embedder pioneer, topped MTEB at release; 7B is GPU-hungry; 2024-era; license: MIT