Qwen2.5 32B Instruct

Permissive

Alibaba · Qwen · Released 2024-09

32BText

A strong permissively-licensed mid-size model that often trades blows with much larger dense models on reasoning benchmarks.

Strengths

  • +Apache 2.0 license
  • +Strong reasoning and coding for its size
  • +128K context window

Limitations

  • -Needs a 24GB+ GPU (or multi-GPU) for good-quality local inference
  • -Slower throughput than 7-14B tiers

License

Apache 2.0

use commercially with attribution niceties

Hardware

wants 24-48GB of VRAM - a 3090/4090-class card or better

Links

Stats

Qwen/Qwen2.5-32B-Instruct— downloads·— likes

via Hugging Face

Related models