Promptriever
An 8B instruction-trained retriever you can prompt like a language model.
About
Promptriever starts from the observation that normal retrievers ignore instructions: give them a query plus a constraint and they retrieve as if the constraint were not there. The paper (arXiv September 2024, by Orion Weller and collaborators) trains a Llama-3.1-8B-based retriever on instruction-augmented retrieval data so the model can be prompted with natural-language guidance the way you would prompt a chat model.
The published weights come from Samaya AI as samaya-ai/promptriever-llama3.1-8b-v1 under Apache 2.0. It is a niche release: the checkpoint has only a few dozen Hugging Face downloads, the paper evaluates instruction-following retrieval on its own benchmark rather than MTEB, and the space has since moved toward reasoning-tuned retrievers. Still, it is the cleanest open example of the promptable-retriever idea.
Specs
Context window
Pricing
Free quota
API access
Weights
Released
Benchmarks
No published MTEB headline; paper evaluates instruction-following retrieval on a purpose-built benchmark (author-reported)
Provider-reported unless a source says otherwise. Benchmarks are marketing-adjacent by nature — treat them as directional.
Use cases
Trust score — 6/10
Scored 6/10 (metrics tier, metrics-v1). Evidence: HF API confirms Apache 2.0 license and Sept 2024 release of the Samaya AI checkpoint; arXiv paper (2409.11136) documents the instruction-training approach. Strengths: novel promptable-retriever idea, permissive license, peer-reviewed-style paper. Docked for: (1) only ~24 HF downloads, so effectively no community validation; (2) no standard benchmark figures to compare against; (3) publisher is a small third party, not the paper's authors.
How the trust score is built
Full methodology →- Adoption · stars/downloads, log-scaled vs peers
- 3.0/10
- Maintenance · recency, releases, issue load
- 5.0/10
- Openness · license + docs
- 9.0/10
- Manual review · sanity read + red-flag check
- 6.0/10
- · HF API confirms Apache 2.0 and Sept 2024 release; arXiv paper 2409.11136
- · Only ~24 HF downloads — effectively no community validation
- · No standard benchmark figures; publisher (Samaya AI) is a third party, not the paper's authors