Qwen 3 32B

Overview

Alibaba's 32B dense flagship with thinking mode, scoring 65.5 on MMLU-Pro and 39.8 on SuperGPQA. The strongest general-purpose Qwen 3 dense model before stepping up to the MoE.

When to pick this model

• You want a single dense model for chat, code, and reasoning on a 48GB-class GPU
• You need multilingual coverage with strong reasoning headroom
• You want one Apache 2.0 model to standardize on for production
• You need 131K context for long-form work

VRAM requirements by quantization

Quantization	VRAM required
Q4_K_M (recommended)	19 GB
Q5_K_M	23 GB
Q8_0	35 GB
FP16 (no quantization)	64 GB

VRAM figures include model weights plus a typical 8k KV cache and ~600 MB runtime overhead (Ollama / llama.cpp baseline). Add headroom for higher context lengths.

Published benchmark scores

Benchmark	Score
MMLU-Pro	65.54
SuperGPQA	39.78

Scores published by the model author or aggregated from public leaderboards. Re-measured monthly by our editorial team.

Strengths

• Strong reasoning with thinking mode enabled
• Solid MMLU-Pro and SuperGPQA scores for its size
• 131K context window
• Apache 2.0 license

Limitations

• QwQ-32B is sharper for pure reasoning tasks
• Verbose thinking traces inflate latency and cost

Architecture & training

Architecture: Dense · GQA · hybrid thinking

Training: Same 36T pre-training as the rest of the Qwen 3 family.

Quick start

ollama run qwen3:32b

Or use the open-source MCP server to query this model from Claude Desktop, Cursor, or any MCP-compatible client.