---
title: tencent/R3-rerank-0.6b
description: "The latest agent skill reranking model at the 0.6B scale. R3-Reranker is the cross-encoder (rerank) stage of R3-Skill's two-stage retriever . Qwen3, 596M parameters."
canonical_url: https://superlinked.com/models/tencent-r3-rerank-0-6b
last_updated: 2026-08-24
---

# tencent/R3-rerank-0.6b

The latest agent skill reranking model at the 0.6B scale. R3-Reranker is the cross-encoder (rerank) stage of R3-Skill's two-stage retriever for query-conditional agent skill retrieval. It scores each (query, skill) pair jointly, paired with R3-Embedding-0.6B for recall.

Source: [tencent/R3-rerank-0.6b on HuggingFace](https://huggingface.co/tencent/R3-rerank-0.6b)
Base model: [Qwen/Qwen3-Reranker-0.6B](https://huggingface.co/Qwen/Qwen3-Reranker-0.6B)

## Overview

| Field | Value |
|-------|-------|
| Architecture | Qwen3 |
| Parameters | 596M |
| Tasks | Score |
| Outputs | Score |
| Max sequence length | 4,096 tokens |
| License | apache-2.0 |
| Inputs | text |

## Benchmarks

### R3Skill (candidates: R3-embedding-0.6b, k=20)

Domain: technology · Task: retrieval · Language: en

Agent skill retrieval: match user requests to the SKILL.md that solves them (Tencent R3 release set)

**Quality:** ndcg at 10: 0.8228

**Performance (L4 b1 c16):** Corpus 25.4K tok/s · Corpus p50 469.9ms · Query 2.2K tok/s · Query p50 469.9ms

[Reference](https://huggingface.co/tencent/R3-embedding-0.6b)

### R3Skill (candidates: R3-embedding-0.6b, k=20, limit 384)

Domain: technology · Task: retrieval · Language: en

Agent skill retrieval: match user requests to the SKILL.md that solves them (Tencent R3 release set)

**Quality:** ndcg at 10: 0.8264

**Performance (L4 b1 c16):** Corpus 25.1K tok/s · Corpus p50 429.0ms · Query 2.1K tok/s · Query p50 429.0ms

**Performance (RTX-PRO-6000-Blackwell-Server-Edition b1 c16):** Corpus 118.0K tok/s · Corpus p50 105.6ms · Query 9.8K tok/s · Query p50 105.6ms

[Reference](https://huggingface.co/tencent/R3-embedding-0.6b)