Qwen2.5 / Qwen3
Information
Qwen2.5 and Qwen3 are model families from Alibaba that are often strong general-purpose choices for both local and
hosted usage. They are widely appreciated for solid reasoning quality, practical multilingual support, Apache 2.0
licensing in many variants, and long context windows that can go up to around 128k tokens depending on the specific
model.
They are available in a broad size range such as 0.5B through 72B, which makes them suitable for both lightweight
local testing and larger GPU-backed deployments.
Why they are commonly recommended
- strong all-round performance for chat, summarization, and general reasoning,
- useful multilingual capabilities,
- open variants that are practical for self-hosted and research workflows,
- and a model size range that works for both small local experiments and larger deployments.
Practical note for Ollama users
Availability in Ollama depends on the current model library and packaging status. Always check the latest supported
tags, quantizations, and names before pulling a model locally.