Qwen · qwen3
Qwen3-32B
Qwen3-32B by Qwen: a 32.8B-parameter dense open-weight model under the apache-2.0 license with tool calling.
5,035,207 downloads · 747 likes · Model card · synced 2026-09-19
Specifications
- Parameters
- 32.8B
- Architecture
- dense · qwen3
- Layers
- 64
- Hidden size
- 5,120
- KV heads · head dim
- 8 · 128
- Vocabulary
- 151,936
- Native dtype
- bfloat16
- Context window
- 40,960 tokens
- Released
- 2025-04-27
- Last modified on Hub
- 2025-07-26
Features & licensing
- License
- apache-2.0
- Commercial use
- Yes
- Modalities
- text
- Tool / function calling
- Yes
- Reasoning mode
- Yes
- Languages
- EN
- Quantised variants on Hub
- GGUF
- Pipeline
- text-generation
Task fit
Editorial scores (0–100) used by the recommendation engine.
- Chat70
- RAG / Q&A70
- Code70
- Summarisation70
- Extraction70
- Agents70
Hardware to run Qwen3-32B
Weights need about 72 GB at FP16, 38 GB at INT8 and 22 GB at INT4. The KV cache adds roughly 262 MB per 1,000 tokens per request. For a reference workload of 5 requests per second with 1,500 input and 300 output tokens, total GPU memory is around 35.6 GB, which fits on 1 × NVIDIA A100 (80 GB).
| Configuration | VRAM | Utilisation | Est. first token | Cloud / month |
|---|---|---|---|---|
| 2 × NVIDIA L4 (24 GB) | 48 GB | 74 % | ~0.5 s | $1,168 |
| 1 × NVIDIA L40S (48 GB) | 48 GB | 74 % | ~0.2 s | $1,387 |
| 1 × NVIDIA A100 (80 GB)recommended | 80 GB | 45 % | ~0.1 s | $2,336 |
| 1 × NVIDIA H100 (80 GB) | 80 GB | 45 % | ~0.0 s | $3,285 |
Estimates only. Run an assessment for your own traffic.
Auto-imported from Hugging Face. Architecture, license and feature data are synced from the model repository; the task-fit scores below are provisional defaults until a curator reviews them.