Model reference · open weights

SkillRouter-Reranker

SkillRouter-Reranker is an open-weight embedding model from pipizhao, listed in the AxForge catalogue. AxForge can bring it up on EU-owned hardware for you on request — with the licence handled where one is required.

Embeddings pipizhao 1 variants 8k downloads/mo
Request this model on EU hardware All served models Not on the shared API today — deployed on request.

About

What SkillRouter-Reranker is

SR-Rank-0.6B SR-Rank-0.6B is a fine-tuned cross-encoder reranker for skill routing. It is designed to score a small candidate set of retrieved skills against a task query and select the single most relevant skill for an LLM agent. Model Summary - Base model: Qwen/Qwen3-Reranker-0.6B - Architecture: causal-LM-style cross-encoder reranker - Input: query + candidate skill text, formatted into the Qwen reranker prompt template - Output: scalar relevance score computed as logit(yes) - logit(no) - Intended use: rerank top-K candidates from a first-stage retriever such as SR-Emb-0.6B This model should be used on a candidate list, not as a standalone retriever over an 80K-scale corpus. Intended Uses Use SR-Rank-0.6B after a first-stage retriever has already narrowed a large corpus to a candidate set, for example: 1. Retrieve top-20 skills with pipizhao/SkillRouter-Embedding-0.6B. 2. Score each candidate with pipizhao/SkillRouter-Reranker-0.6B. 3. Sort by yes - no score. 4. Take the top-1 or top-N skills for downstream planning or execution. This model is not intended for free-form generation, chat, or long-document retrieval without candidate pruning. How to Use Citation If you use this model, please cite the SkillRouter paper once the preprint is public.

Summarised from the published model card. Read the full card on the HuggingFace links below.

Specifications

What it is

Makerpipizhao
TypeEmbedding models
Parameters (lead)596M
Context40k tokens
Variants1
Runs withtransformers
Based onQwen/Qwen3-Reranker-0.6B
Released2026-03-26
Popularity8k downloads / month
Likes6
LicenceOpen weights

How it works

How embedding models work

Your textsentence / documentEncodermaps meaningVectorlist of numbersAn embedding model turns text into a vector, so similar meanings sit close together — the basis of search and RAG.

Variants

Sizes & precisions

Open weights ship in several sizes and precisions. One page, all the variants — pick the one that fits your GPU. VRAM figures are estimates from model size.

VariantParamsPrecisionVRAMFits 16 GBWeights
SkillRouter-Reranker-0.6B596MBF16~1.4 GBWeights ↗

Using it via the API

Call it like any OpenAI endpoint

Once AxForge deploys skillrouter-reranker for you, it answers on the OpenAI-compatible API — the same base URL and keys as every other model. (skillrouter-reranker below is illustrative; you get the exact model name on deployment.)

$ curl -sS https://api.axforge.ai/v1/embeddings \
  -H "Authorization: Bearer $AXFORGE_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{"model":"skillrouter-reranker","input":"text to embed"}'

Details

Languages, data & research

Languages

en

Tags

transformers safetensors qwen3 text-generation reranking cross-encoder agent-tools skill-routing listwise-learning text-ranking en endpoints_compatible

Papers

Licence

Open weights

Open weights under apache-2.0 — commercial use is permitted. Deploy it on AxForge EU hardware on request. Read the licence ↗

Sources

Weights & code

Want SkillRouter-Reranker on EU-owned hardware?

Request this model on EU hardware See what’s served now

Explore

More embedding models

© 2026 AxForge · EU-hosted AI infrastructure Pricing Docs Trust Privacy Terms