Qwen3 Technical Report
By An Yang, Anfeng Li, Baosong Yang, Beichen Zhang, Binyuan Hui, Bo Zheng, Bowen Yu, Chang Gao, Chengen Huang, Chenxu Lv and 48 more. 2025
6,644 citations recorded by the source corpus.
In this work, we present Qwen3, the latest version of the Qwen model family. Qwen3 comprises a series of large language models (LLMs) designed to advance performance, efficiency, and multilingual capabilities. The Qwen3 series includes models of both dense and Mixture-of-Expert (MoE) architectures, with parameter scales ranging from 0.6 to 235 billion. A key innovation in Qwen3 is the integration of thinking mode (for complex, multi-step reasoning) and non-thinking mode (for rapid, context-driven responses) into a unified framework. This eliminates the need to switch between different models--such as chat-optimized models (e.g., GPT-4o) and dedicated reasoning models (e.g., QwQ-32B)--and enables dynamic mode switching based on user queries or chat templates. Meanwhile, Qwen3 introduces a thinking budget mechanism, allowing users to allocate computational resources adaptively during inference, thereby balancing latency and performance based on task complexity. Moreover, by leveraging the knowledge from the flagship models, we significantly reduce the computational resources required to build smaller-scale models, while ensuring their highly competitive performance. Empirical evaluations demonstrate that Qwen3 achieves state-of-the-art results across diverse benchmarks, including tasks in code generation, mathematical reasoning, agent tasks, etc., competitive against larger MoE models and proprietary models. Compared to its predecessor Qwen2.5, Qwen3 expands multilingual support from 29 to 119 languages and dialects, enhancing global accessibility through improved cross-lingual understanding and generation capabilities. To facilitate reproducibility and community-driven research and development, all Qwen3 models are publicly accessible under Apache 2.0.
Subfields: Artificial Intelligence, Computer Vision and Pattern Recognition, Health Informatics
View Paper: https://www.semanticscholar.org/paper/d2d84d56f730f81d276a02b48d5d44db5bde0b4a
How researchers see this paper
The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.
In Artificial Intelligence
In Computer Vision and Pattern Recognition
Citations in the corpus
- Qwen3-Omni Technical Report (2025, positive): “The second stage adopts the Strong-to-Weak Distillation pipeline as described in Qwen3 (Yang et al., 2025a) to further improve model performance.”
- Qwen3 Embedding: Advancing Text Embedding and Reranking Through Foundation Models (2025, neutral): “…paradigms such as Retrieval-Augmented Generation (RAG) and agent systems, driven by the advancement of large language models (e.g., Qwen3 (Yang et al., 2025), GPT-4o (Hurst et al., 2024)), have introduced new requirements and challenges for text embedding and reranking, both in terms of…”
- Jailbreak and Guard Aligned Language Models With Only Few In-Context Demonstrations (2023, neutral): “5 [69], Llama-2 [23], Llama-3 [23], QWen-2 [72], QWen-3 [73], Mistral-2 [74]) and 4 closed-sourced models ( GPT-4 [22], Claude-2 [75], Gemini-2.”
- GLM-4.5: Agentic, Reasoning, and Coding (ARC) Foundation Models (2025, neutral): “The new models outperform existing open-source LLM models [13; 34; 47] across the board, with significant gains in agentic, reasoning, and coding tasks.”
- OntoLearner: A Modular Python Library for Ontology Learning with Large Language Models (2026, unknown): “From the Qwen family (Yang et al. 2025), we evaluated base, instruction-tuned, and reasoning-oriented variants, including Qwen3-0.”
- TAGGRAPH: Tag-Augmented Graphs for Graph Retrieval of Agent Persistent Histories (2026, unknown): “Section 5.3 attributes part of Gemma’s advantage to a more convergent tag vocabulary.”
- Q-TIE: A Lightweight and Generalizable Re-ranking Framework for Temporal Information Retrieval (2026, unknown): “6B (Yang et al., 2025), chosen to balance extraction accuracy with inference efficiency.”
- Recall Before Rerank: Benchmarking Deep Learning Models for Large-Scale Code-to-Code Retrieval (2026, unknown): “…bi-encoders (e.g., CodeRankEmbed [30]), encoder-decoders (e.g., CodeT5 [33]), and decoder-only LLMs in embedding mode (e.g., Code Llama [28], Qwen3 [35]); (ii) model scale : favouring the base variant of each family; (iii) provenance : open-source weights, code-specialised training, and varied…”
- GraphRAG on Consumer Hardware: Benchmarking Local LLMs for Healthcare EHR Schema Retrieval (2026, unknown): “We implement the Microsoft GraphRAG pipeline on real-world Epic Clarity schema documentation and benchmark four open-source models such as Llama 3.1 (8B), Mistral (7B), Qwen 2.5 (7B), and Phi-4-mini (3.8B) on a single consumer GPU [15, 13, 16, 14].”
- ASTRA-QA: A Benchmark for Abstract Question Answering over Documents (2026, unknown): “All RAG methods use Qwen3-8B [34] as the LLM backbone and nomic-embed-text [35] as the embedding model.”