Qwen3
Alibaba's Qwen3 family of open-weight large language models, in dense and mixture-of-experts variants.
Liked as open-weights models you can run at home12, from small quantized builds to promising MoE variants,3456 and as robust agents with tools like opencode.7 No single complaint dominates: one user says local use can take longer than big cloud models,8 another prefers Qwen 2.5 Coder for executing code, and heavily quantized versions need explicit prompting to use tools.910
Why people recommend it
Runs locally · 3
Open weights let users run it at home or cheaply through providers instead of relying on cloud models.
“open weights models like Qwen3 and kimik2 exceeded their capability and you can run them at home”
“QWEN3* models appear to be very robust and can be prompted with opencode using basic bash scripts and robust plugins”
Show 1 more
“the heavily quantized Qwen3 models I can cram onto my 3060 Ti”
Small model options · 2
Small and heavily quantized versions fit modest hardware like a 3060 Ti.
“qwen3 1.7b- q4_k_m is your best best for that size”
“the heavily quantized Qwen3 models I can cram onto my 3060 Ti”
MoE models · 2
The mixture-of-experts variants are described as promising and a good fit.
“recent qwen3.x MoE models look very promising”
“just checked it out, has mixture of experts, mit fit my use case!”
Agent robustness · 1
Seen as robust when driven by opencode with bash scripts, plugins and prompts.
“QWEN3* models appear to be very robust and can be prompted with opencode using basic bash scripts and robust plugins”
Fast and reasoning · 1
One user relies on the fast mode for quick questions and the reasoning mode for structured research.
“I tend to rely on Qwen3 (fast) for quick questions and Qwen3 (reasoning) when I want a more structured approach”
Upgrade over 2.5 · 1
Widely regarded as a significant upgrade over Qwen2.5.
“Yet Qwen3 was widely regarded as a significant upgrade to Qwen2.5.”
Caveats and dissent
Slower locally · 1
Running these models instead of larger cloud models might take longer.
“QWEN3* models appear to be very robust and can be prompted with opencode using basic bash scripts and robust plugins”
Coding execution · 1
One user finds Qwen 2.5 Coder 30B better than Qwen3 for executing coding plans.
“Qwen3-235B - between Qwen 2.5 and Sonet”
Tool use prompting · 1
Heavily quantized versions were more reliable at using tools when the system prompt told them to.
“the heavily quantized Qwen3 models I can cram onto my 3060 Ti”
Often compared with
Who recommends it in their own words
Threads where it came up
- Ask HN: Best local LLM under 2B paramater and consuming RAM less than 3gb 2 mentions
- Ask HN: What's your AI coding set up? 1 mention
- Running local models is good now 1 mention
- Show HN: Learn from 30 historical figures, open source, nonprofit, self-hosted 1 mention
- The New Linux Kernel AI Bot Uncovering Bugs Is a Local LLM on Framework Desktop 1 mention
- Ask HN: What Online LLM / Chat do you use? 1 mention
Also written as Qwen3 1.7B, Qwen3-235B, Qwen3 235B. Data: qwen3.json. Quotes are verbatim; the verdict and theme labels are written by a model from these comments, and every clause links to its sources.
All comments 10
“recent qwen3.x MoE models look very promising”
“just checked it out, has mixture of experts, mit fit my use case!”
“qwen3 1.7b- q4_k_m is your best best for that size”
“the heavily quantized Qwen3 models I can cram onto my 3060 Ti”
“It is the base model, Qwen3 235B for the free tier”
“QWEN3* models appear to be very robust and can be prompted with opencode using basic bash scripts and robust plugins”
“I tend to rely on Qwen3 (fast) for quick questions and Qwen3 (reasoning) when I want a more structured approach”
“Yet Qwen3 was widely regarded as a significant upgrade to Qwen2.5.”
“Qwen3-235B - between Qwen 2.5 and Sonet”
“open weights models like Qwen3 and kimik2 exceeded their capability and you can run them at home”