Qwen 3

What Qwen 3 is for

Qwen 3 is a reasoning model: it spends output tokens working through a problem before answering, which makes it slower per useful word and better on problems that need the working shown. Alibaba released it in 2025-04 with 8 sizes (0.6B, 1.7B, 4B, 8B, 14B, 30B, 32B, 235B) under Apache 2.0 (most sizes), with a 128K context window. Budget for the extra tokens the thinking phase consumes. The catalogue records its strength as best all-round local LLM in 2025-2026.

Sizes and memory

The 8B needs about 5 GB of weights at Q4_K_M and the 235B about 130 GB, a spread of 125 GB across the family. The catalogue's default pick is 30B. Recommended VRAM adds roughly half again on top of the weights for the KV cache and the runtime.

Size	Weights at Q4_K_M	Recommended VRAM	Smallest card in our guide that holds it
8B	5 GB	8 GB	NVIDIA Jetson Orin Nano 8GB
14B	9 GB	14 GB	NVIDIA RTX 3060 12GB
30B	18 GB	27 GB	NVIDIA RTX 3090
32B	20 GB	30 GB	NVIDIA RTX 3090
235B	130 GB	195 GB	Apple M3 Ultra

Hardware that runs Qwen 3

Every one of the 19 cards with dedicated memory in our hardware guide holds at least one size of Qwen 3 at Q4_K_M. The cheapest that does is the NVIDIA Jetson Orin Nano 8GB at about $199, running the 8B.

The largest size any of them holds is the 235B at 130 GB, on the Apple M3 Ultra (192 GB at 800 GB/s). Bandwidth, not capacity, sets the speed once a size fits, so a card further down the list with faster memory will generate more tokens per second on the same weights than a larger, slower one.

What people use it for

The catalogue lists 5 recorded uses for Qwen 3: chat, code, reasoning, RAG, agentic. It is tagged chat, thinking, MoE, tools, multilingual, agentic.

Where Qwen 3 sits in the Qwen family

Our catalogue holds 5 Qwen entries. Ordered by release date, Qwen 3 (2025-04) is the 3rd of them, and Qwen 3 VL (2025-10) is newer. Qwen 2.5 and Qwen 2.5-Coder carry the same recorded memory footprint, so choosing between those and this one is a question of behaviour rather than of hardware: Qwen 3 is the one recorded here for best all-round local LLM in 2025-2026.

The Apache 2.0 (most sizes) licence puts no usage ceiling on it, which matters if the thing you are building has users. That is the practical difference between this and a vendor community licence.

Context window: 128K

128K tokens is roughly 96,000 words at the usual ratio, which is a small repository or a whole book. Two cautions come with a window this size: the KV cache at full length can rival the weights for memory, and models rarely use the far end of their advertised context as well as they use the near end. Treat it as headroom, not as a promise.

How to run Qwen 3 locally

Our catalogue does not record registry tags, so look the current tag up in the Ollama library or on Hugging Face before pasting these. The commands below are the shape of the workflow, not a copy-and-paste recipe.

Option 1: Ollama (simplest)

# find the tag for Qwen 3 at ollama.com/library
ollama run <tag>

Option 2: Mullama (production)

mullama pull <tag>
mullama run <tag>

Option 3: llama.cpp (CLI)

# download a GGUF from Hugging Face, searching for "Qwen 3 GGUF"
./llama-cli -m model.Q4_K_M.gguf -p "Hello, AI!"

Option 4: Python with Mullama or llama-cpp-python

from mullama import Model, Context
model = Model.load("model.Q4_K_M.gguf", n_gpu_layers=99)
ctx = Context(model, n_ctx=4096)
print(ctx.generate("Hello, AI!", 256))

Sources

Model record from src/data/models.json (2026-06-29); card capacities from src/data/gpus.json (2026-06-29). Upstream: Curated from ollama.com/library + community benchmarks + paperswithcode + huggingface.