Skip to content

Home Leaderboards News Models Tools

Loading

FitStack starts from the job. Recipes belong to their authors.

Stacks Examples Glossary Compare Today Benchmarks Status Radar

A guide, not a directory

FitStack starts from the job, not from a logo wall. You say what you are trying to do, what machine you have, and whether the data may leave the building. You get two or three alternatives: the easy cloud path, a builder-in-the-middle, and a local stack when the hardware can carry it.

We do not keep our own install tutorials. The steps live with the people who maintain them — MiaAI-Lab’s Spark recipes, NVIDIA’s playbooks and NemoClaw, OpenClaw, and xAI’s Grok Build surfaces. A field note here is a reason to click, plus an official one-liner when the authors published one.

Cloud rates come from OpenRouter, stamped with the date they were fetched. Street prices, local fits, and electricity are authored ballparks, not live listings. Beginner scores are our own judgement.

Recipes belong to their authors. FitStack does not maintain the install scripts.

Stacks Examples Glossary Compare Today Benchmarks Status Radar

Home
Leaderboards
News
Models
Tools

Models Leaderboards

Qwen3 4B

local · Updated 56 min ago

No price history is stored for this model.

Scores

No score stored.

Skill.: No number
Coding.: No number
Tools.: No number
Context.: No number

No coding rank is stored for this model.

Cost

No hosted list cost for this model.

On your machine

Qwen3 4B at Q4_K_M needs about 3 GB for the weights.

How to use it

No hosted model id is stored, so there is no API call to copy.

llama.cpp downloads Qwen3-4B-Q4_K_M.gguf from Qwen/Qwen3-4B-GGUF. About 3 GB for the weights at Q4_K_M.

llama.cpp

llama-cli -hf Qwen/Qwen3-4B-GGUF:Q4_K_M

Qwen3-4B-Q4_K_M.gguf

No Ollama tag is stored for Qwen3 4B, so there is no ollama run line to copy.

Ollama

Latest stories

No story in the feed names this model.

Try the recipe See it on the local board