Workflows/Analyze images locally with Qwen3.5

Images / WORKFLOW

Analyze images locally with Qwen3.5

Set up Qwen3.5 for local image understanding through a receipt-extraction example. Choose the lighter 2B or more capable 4B package, upload a receipt, and compare the extracted fields with the original. The supplied app is a dedicated receipt reader, not a general image-chat interface.

EstimatedSource checked 9/29/2026

LOCAL

Freshness 41 out of 100. Repository age: 230 days old, Feb 27, 2026. Measures age, not quality.

Popularity 91 out of 100. 7,944,995 downloads in 30 days; 1,073 lifetime likes. Checked Oct 15, 2026. Measures adoption, not quality.

What do these scores mean?

Two ways to compare models. Neither score measures answer quality, benchmark performance, or whether a model fits your machine.

Freshness · release age

A new release starts at 100. The score halves every 180 days: 100 × 2^(−age in days / 180). When the release date is unavailable, a repository date is a fallback—not proof of when the model was released. Missing dates stay unscored.

Check the date source

Popularity · adoption

Downloads over the past 30 days carry 80% of the score; lifetime likes carry 20%. Both use a logarithmic scale, capped at 10 million downloads and 100,000 likes. If likes are unavailable, downloads carry all the weight. Missing download statistics stay unscored.

round(80 × min(1, log10(1 + downloads) / 7) + 20 × min(1, log10(1 + likes) / 5))

Check the statistics source

Scores as of Oct 15, 2026. Statistics checked Oct 15, 2026. Statistics are marked for refresh after 72 hours.Full scoring method

Before you begin

Hardware

Starting estimates for one receipt at 4K context. Memory applies to the complete Ollama vision request, not weights alone.

Machine	Qwen3.5 2B Q4	Qwen3.5 4B Q4
NVIDIA GPU · 4 GB VRAM + 16 GB RAM	Start here; limited GPU headroom	Use 2B first or allow CPU offload
NVIDIA GPU · 8 GB VRAM + 16 GB RAM	Comfortable starting estimate	Suitable starting estimate for one image
AMD GPU · 8 GB VRAM + 16 GB RAM	Use a supported Ollama GPU backend	Use a supported Ollama GPU backend
Intel GPU · 8 GB VRAM + 16 GB RAM	Vulkan when supported; CPU fallback otherwise	Vulkan when supported; CPU fallback otherwise
Apple Silicon · 8 GB unified memory	Smaller option; close memory-heavy apps	Possible but tight; 16 GB is preferable
Apple Silicon · 16 GB unified memory	Comfortable starting estimate	Recommended Mac starting point
CPU only · 16 GB RAM	Usable CPU route; wait longer per image	Usable CPU route; slower than accelerated inference

No speed claims are implied. Check the exact GPU/OS against Ollama’s supported backends; insufficient VRAM may offload to system RAM.

Context window

Machine	Qwen3.5 2B Q4	Qwen3.5 4B Q4
GPU · 8 GB / Mac · 16 GB unified	Start at 4,096 tokens; one receipt image	Start at 4,096 tokens; one receipt image

The app bounds each request. Increasing context or the number/resolution of images raises memory use; single-page extraction rarely needs a very long conversation.

LOCAL IMAGE UNDERSTANDINGSOURCE-LINKED SETUP

Analyze images locally with Qwen3.5

Explore local image understanding with Qwen3.5 using a receipt-extraction example. The supplied app reads a receipt image and saves its merchant, date, total and currency as JSON; it is a dedicated receipt reader. Choose the lighter 2B or more capable 4B package and check every extracted field against the original.

Choose a package Run its commands Check the result

BEFORE YOU BEGIN

Choose your package

Select the weights you want. The instructions, downloads and settings change with your choice.

Save your machine in My Hardware to get an automatic starting choice. You can always choose any package yourself.

CURRENT SETUP

Qwen3.5 2B Q4 · official

Estimated for one receipt image and a 4,096-token request. The app sets that limit explicitly; larger images need more working memory. Check Ollama GPU support for the exact card.

Freshness unavailable. Date pending.

Popularity unavailable. Stats pending.

What do these scores mean?

Two ways to compare models. Neither score measures answer quality, benchmark performance, or whether a model fits your machine.

Freshness · release age

Check the date source

Popularity · adoption

round(80 × min(1, log10(1 + downloads) / 7) + 20 × min(1, log10(1 + likes) / 5))

Scores as of Oct 15, 2026. Statistics check pending. Statistics are marked for refresh after 72 hours. These statistics are due for a refresh.Full scoring method

View weight source

THE WORKFLOW

Set up Qwen3.5 2B Q4 on your machine

Choose where you want to run it. The instructions below update to match your selection.

CHOOSE YOUR SETUPWindows

NVIDIA uses CUDA; AMD/Intel acceleration depends on the exact card and supported ROCm/Vulkan driver. CPU fallback is available with sufficient RAM.

Install and start Ollama on Windows

Use the linked native installer, finish setup, and open the Ollama app. Keep it running.

Install Ollama GPU and driver support

Download the 2B vision model

This downloads the selected model to Ollama once. It does not start a terminal chat; the browser app will send the image and request the structured fields.

Exact package

ollama pull qwen3.5:2b

Install uv · PowerShell

Install uv from its official instructions, then open a new PowerShell window. The commands below create an isolated Python 3.11 environment; there is no need to change your system Python.

uv installation

powershell -ExecutionPolicy ByPass -c "irm https://astral.sh/uv/install.ps1 | iex"

Create the project and install its dependencies

Use a new folder named yourunai-receipts in your home directory. These commands run in PowerShell. Leave this window in the project folder for the remaining commands.

uv virtual environments

New-Item -ItemType Directory -Force -Path "$env:USERPROFILE\yourunai-receipts" | Out-Null
Set-Location "$env:USERPROFILE\yourunai-receipts"
uv venv --python 3.11
uv pip install --python .\.venv\Scripts\python.exe ollama==0.6.2 gradio==6.28.0 pillow==12.3.0

Download the complete starter app

Right-click the download link below and choose Save Link As. Save the .py file directly in the project folder created above; keep its .py extension, not .txt. This complete YouRunAI starter creates the browser interface and output folders. You can inspect it in a text editor. Model weights download separately when the runtime needs them.

Download receipt-reader.py

Launch the receipt interface

Run this in PowerShell and keep the process open. Open http://127.0.0.1:7860. The app creates a readable sample-receipt.png and sends images only to Ollama at 127.0.0.1:11434. Each request uses 4,096 context tokens with a 1,024-token answer limit.

.\.venv\Scripts\python.exe receipt-reader.py --model qwen3.5:2b

Extract the supplied receipt

The sample image is already selected. Click Extract receipt to run vision inference. Compare the JSON with the image: merchant CEDAR CAFE, date 2026-09-28, total 12, currency USD. A structurally valid JSON response can still contain a reading error, so check the actual fields.

Upload your own receipt and save the JSON

Replace the image with a clear PNG or JPEG photo of your own receipt, then click Extract receipt. Keep the whole receipt legible and upright. Download the JSON using the file button; every result also gets a separate timestamped file under outputs/. PDF pages must first be exported as PNG/JPEG. Null means the field was not legible; compare it with the image before using the result.

Stop and return to the same project later

Press Ctrl+C in the running PowerShell window to stop the app. Your files stay in yourunai-receipts. Next time run the commands below, then reopen http://127.0.0.1:7860. Dependencies and downloaded weights are reused. Keep this address local; there is no public sharing link.

Set-Location "$env:USERPROFILE\yourunai-receipts"
.\.venv\Scripts\python.exe receipt-reader.py --model qwen3.5:2b

SUCCESS CHECK · Qwen3.5 2B Q4

A receipt you can verify and a file you can keep

The browser shows the original image and four extracted fields. The sample totals match the visible receipt, and a timestamped JSON file appears in outputs/. Your own receipt is checked the same way.

Back to setup steps

When it doesn’t go to plan

The browser cannot connect

Keep the Python process open and wait until it prints the local URL. If another app uses that port, stop that app or run these examples one at a time.

The first launch appears slow

The first run downloads the publisher model to the Hugging Face cache. Watch the Terminal/PowerShell progress and allow the download to finish; later starts reuse it.

A dependency is missing

Run the installation command from this guide again in the project folder. Launch with the .venv Python path shown here so you use the same environment.

The model runs out of memory

Close memory-heavy apps and other models. Keep the supplied short input limits and batch size. The small-model CPU route needs system RAM, not GPU VRAM.

Ollama connection refused

Open the Ollama app on Windows/Mac, or start its Linux service. Keep one server running on port 11434, then retry Extract receipt.

Model not found or vision unsupported

Pull the exact qwen3.5:2b or qwen3.5:4b tag shown here. Update Ollama and restart its server. A text-only substitute cannot read the image.

The total or merchant is wrong

Use a sharp, upright, tightly cropped image and compare every field with the receipt. Try the 4B package if the smaller model misses clear text; do not treat valid JSON as proof of accuracy.

REFERENCE LIBRARY

Sources and files

Original instructions, model files, and compatibility notes behind this setup.

7 SOURCES

01Qwen3.5 2B packageollama.com 02Qwen3.5 4B packageollama.com 03Ollama vision inputsdocs.ollama.com 04Ollama structured outputsdocs.ollama.com 05Ollama GPU supportdocs.ollama.com 06uv installationdocs.astral.sh 07Receipt app sourceyourunai.com

The model behind this workflow

Qwen3.5 · 4B

Keep exploring

Text

SETUP GUIDE

Qwen3.5 · 9B private chat setup

Source-backed package and platform-route plan for private local chat with Qwen3.5 · 9B. Preserves the official Ollama 9B package, a smaller 4B family fallback, and a clearly labeled community uncensored GGUF option.

Freshness 41 out of 100. Release age: 230 days old, Feb 27, 2026. Measures age, not quality.

Build a private coding assistant with Qwen3.8 27B

Run Qwen3.8 27B with Ollama, connect it to Cline in VS Code, and make one reviewable code change in a tiny local project.

Freshness 76 out of 100. Repository age: 71 days old, Aug 5, 2026. Measures age, not quality.

Arena Open Source Web coding: rank 10 of 55. Ranked model: qwen3.8-27b. Original model ranking, not a benchmark of this package. Export dated Oct 13, 2026. Source checked Oct 15, 2026. Human preference ranking, not hardware suitability.+3

SOFTWAREOllama + Visual Studio Code + Cline + Python

Create images locally with FLUX.2 Klein

Generate images from text prompts in ComfyUI with the fast distilled or more adjustable base 4B model. Save your results and reuse the visual workflow.

Freshness 35 out of 100. Release age: 273 days old, Jan 15, 2026. Measures age, not quality.

SOFTWAREComfyUI

LOCALRENTED GPU

Verified source