Chat privately with Qwen3.5 4B in the Ollama app on Windows or Mac, or in its interactive Terminal on Linux. Start the exact 3.4 GB package at 4K context and check a follow-up message.
Estimated capacity for the exact package at a 4K starting context.
| Machine | Qwen3.5 · 4B |
|---|---|
| NVIDIA GPU 8 GB+ VRAM | Estimate: 8 GB VRAM min at short context; 16 GB recommended. Ollama NVIDIA route documented; verify driver. |
| AMD GPU 8 GB+ VRAM | Estimate: 8 GB VRAM min short context; 16 GB recommended. Ollama AMD ROCm/Vulkan path conditional; verify. |
| Intel Arc GPU 16 GB+ VRAM | 16 GB VRAM estimate. Intel Vulkan needs a supported driver/backend; confirm acceleration or use 16 GB system RAM. |
| Apple Silicon unified memory 16 GB+ | Estimate: 16 GB unified memory starting point; 3.4 GB package plus overhead. Ollama macOS Apple M-series route. |
| CPU system RAM 16 GB+ | Estimate: 16 GB RAM starting point; 3.4 GB package plus overhead. Ollama CPU route documented; may be slow. |
Capacity estimates are not speed tests. GPU support depends on the driver/backend. Keep the context at 4K until the first chat works with memory to spare.
Start the selected package at 4K; increase only with available memory.
| Machine | Qwen3.5 · 4B |
|---|---|
| NVIDIA GPU 8 GB+ VRAM | 4096 tokens start; Ollama default 4k under 24 GB VRAM. Publisher max 262144; increase only with headroom. |
| AMD GPU 8 GB+ VRAM | 4096 tokens start; Ollama AMD ROCm/Vulkan conditional. Publisher max 262144; increase only with VRAM headroom. |
| Intel Arc GPU 16 GB+ VRAM | 4096 tokens start with a working Vulkan backend; otherwise use the CPU route with system RAM. |
| Apple Silicon unified memory 16 GB+ | 4096 tokens start; publisher max 262144, but begin short and increase with unified-memory headroom. |
| CPU system RAM 16 GB+ | 4096 tokens start; Ollama CPU route. Publisher max 262144; increase only with RAM headroom. |
The starting context is a practical memory estimate. A larger model context ceiling is not a recommendation for every machine.
Chat privately with Qwen3.5 4B in the Ollama app on Windows or Mac, or in its interactive Terminal on Linux. Start the exact 3.4 GB package at 4K context and check a follow-up message.
This setup uses the package documented for this task. Review its source and supported platforms before starting.
Save your machine in My Hardware to get an automatic starting choice. You can always choose any package yourself.
Estimated for a 4K starting context: 8 GB GPU memory minimum, 16 GB recommended. Start with 16 GB unified memory on Mac or 16 GB system RAM for CPU use. The 3.4 GB download needs extra runtime/context memory; longer context needs more headroom. GPU offload also needs an estimated 16 GB system RAM. These are capacity estimates, not speed tests.
Choose where you want to run it. The instructions below update to match your selection.
Apple Silicon uses GPU acceleration; Intel Macs use the CPU.
Already have the current Ollama app? Open it and skip installation. Otherwise download the Mac app, drag it into Applications and open it on macOS 14 or newer. The selected package downloads 3.4 GB; leave additional disk space for Ollama and your files. Use the memory estimate above and keep Ollama open.
Download Ollama for Mac Installation requirementsStart a new chat in Ollama. Open the model picker, search for qwen3.5:4b and select that exact name, including the tag after the colon. Allow the 3.4 GB download to finish when first used. Confirm the selected model remains qwen3.5:4b before sending a message. The app manages the downloaded model files.
Exact selected package Ollama desktop chatIn Ollama Settings, move the context-length slider to 4K (4096 tokens). Start a fresh conversation at this setting. This guide's memory estimates use 4K. Longer context consumes more memory; try 8K only after the initial chat works with memory to spare.
Ollama context settingsIn the chat box, type: “Remember that my test word is cedar. What is 7 plus 5?” Send it and check that the answer is 12. In the same chat, send the follow-up: “What test word did I ask you to remember?” Check that it says cedar, then continue with your own questions.
Reopen this chat from the app's conversation list to continue. Start a new chat for a fresh topic or another model. Wait for the active reply before closing the app. Quit Ollama from its menu-bar icon to stop the local service; reopen it from Applications next time. The model stays downloaded.
With qwen3.5:4b selected, check that 7 plus 5 returns 12 and the follow-up recalls cedar. Continue in that conversation. This tests basic interaction and continuity, not general answer accuracy.
This is a source-linked setup, not a YouRunAI hardware test. Confirm your exact runtime version, package, and output before relying on it.
Update Ollama, reopen the app and select qwen3.5:4b again. Check the network and free disk space; the weights alone need 3.4 GB.
Close memory-heavy apps and set 4K context in Settings or /set parameter num_ctx 4096 inside the Linux chat. From a separate Terminal or PowerShell window run ollama ps while a reply is active to inspect CPU/GPU use. Consult the linked GPU documentation for your driver/backend.
Start a new chat and repeat the short two-message check. Long conversations can exceed the context; shorten them or increase context only when memory allows.
Original instructions, model files, and compatibility notes behind this 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.
Run Qwen3.8 27B with Ollama, connect it to Cline in VS Code, and make one reviewable code change in a tiny local project.
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.