LLM EXPLORER 60,783 MODELS INDEXED

QwQ 32B Preview by Qwen

By Qwen · 42030 downloads
HF Repo ↗ Copy URL

QwQ 32B Preview is an open-source language model by Qwen. Features: 32b LLM, VRAM: 65.5GB, Context: 32K, License: apache-2.0, Instruction-Based, LLM Explorer Score: 0.48, ELO: 1168.

  Arxiv:2407.10671 Base model:finetune:qwen/qwen2... Base model:qwen/qwen2.5-32b-in...   Chat   Conversational   En   Endpoints compatible   Eval-results   Instruct   Qwen2   Region:us   Safetensors   Sharded   Tensorflow
Model Card on HF 🤗: https://huggingface.co/Qwen/QwQ-32B-Preview 

QwQ 32B Preview Benchmarks

LMSys ELO 1168 vs 1272 (so35) -8.2%
nn.n% — How the model compares to the reference models: Anthropic Sonnet 3.5 ("so35"), GPT-4o ("gpt4o") or GPT-4 ("gpt4").

QwQ 32B Preview Parameters and Internals

Model Type 
Causal Language Models
Use Cases 
Limitations:
Language Mixing and Code-Switching, Recursive Reasoning Loops, Safety and Ethical Considerations, Performance and Benchmark Limitations
Additional Notes 
As a preview release, it demonstrates promising analytical abilities while having several important limitations.
Supported Languages 
en (Native)
Training Details 
Context Length:
32768
Model Architecture:
transformers with RoPE, SwiGLU, RMSNorm, and Attention QKV bias
Responsible Ai Considerations 
Mitigation Strategies:
Requires enhanced safety measures to ensure reliable and secure performance.
Input Output 
Input Format:
text input with chat template
Accepted Modalities:
text
Output Format:
text
LLM NameQwQ 32B Preview
Repository 🤗https://huggingface.co/Qwen/QwQ-32B-Preview 
Base Model(s)  Qwen/Qwen2.5-32B-Instruct   Qwen/Qwen2.5-32B-Instruct
Model Size32b
Required VRAM65.5 GB
Updated2026-07-27
MaintainerQwen
Model Typeqwen2
Instruction-BasedYes
Model Files  3.9 GB: 1-of-17   3.9 GB: 2-of-17   3.9 GB: 3-of-17   3.9 GB: 4-of-17   3.9 GB: 5-of-17   3.9 GB: 6-of-17   3.9 GB: 7-of-17   3.9 GB: 8-of-17   3.9 GB: 9-of-17   3.9 GB: 10-of-17   3.9 GB: 11-of-17   3.9 GB: 12-of-17   3.9 GB: 13-of-17   3.9 GB: 14-of-17   3.9 GB: 15-of-17   3.9 GB: 16-of-17   3.1 GB: 17-of-17
Supported Languagesen
Model ArchitectureQwen2ForCausalLM
Licenseapache-2.0
Context Length32768
Model Max Length32768
Transformers Version4.43.1
Tokenizer ClassQwen2Tokenizer
Padding Token<|endoftext|>
Vocabulary Size152064
Torch Data Typebfloat16
Errorsreplace

Quantized Models of the QwQ 32B Preview

Model
Likes
Downloads
VRAM
PathfinderAI01765 GB
...eview Gptqmodel 4bit Vortex V2161315 GB
QwQ 32B Preview 6bit43226 GB
QwQ 32B Preview GPTQ 4bit330616 GB
...eview Gptqmodel 4bit Vortex V1512616 GB
QwQ 32B Preview Bnb 4bit412819 GB
QwQ 32B Preview AWQ243919 GB
...Q 32B Preview Unsloth Bnb 4bit191123 GB
Qwen QwQ 32B Preview MLX 8bit43034 GB
QwQ 32B Preview 8bit61634 GB

Best Alternatives to QwQ 32B Preview

Best Alternatives
Context / RAM
Downloads
Likes
...y Qwen2.5coder 32B V24.1q 200K195K / 65.8 GB92
...wen2.5 32B Inst BaseMerge TIES128K / 65.8 GB4517
...wen2.5 32B Inst BaseMerge TIES128K / 65.8 GB644
Franqwenstein 35B128K / 69.8 GB58
ELYZA Thinking 1.0 Qwen 32B128K / 65.8 GB2310
Archaeo 32B KTO128K / 65.8 GB164
Hamanasu Magnum QwQ 32B128K / 65.8 GB2959
Qwen2.5 32B Gokgok Step3128K / 65.7 GB60
Openthaigpt R1 32B Instruct128K / 65.8 GB338
Qwen2.5 32B Dark Days Stage2128K / 65.8 GB90