Qwen Image Max

Qwen Image Max is the premium tier of Alibaba's Qwen image model - the same family as the cheap Qwen Image text-to-image node, but a step up in quality at roughly three and a half times the price ($0.07 vs $0.02 per image). If you've hit the ceiling of the standard tier on detail, composition, or prompt adherence, this is the upgrade that lives in the same pack, same API, same wiring.

It runs through the WaveSpeed AI API via the ERPK Collection, the community pack wrapping the service for ComfyUI. No local model, billed per image.

How it works

Type a prompt (max 800 chars - tighter than the standard tier's), pick width and height, and the node sends it to WaveSpeed's Qwen Image Max endpoint. When the render lands, the node downloads it and hands you a ComfyUI image IMAGE tensor, ready for Save or downstream nodes. The seed is a real seed (-1 for random), so you can actually lock a look and explore around it.

The honest framing: 800 characters is plenty for a good prompt, but this is a model that rewards deliberate prompting, not rambling. Give it subject, style, composition, and lighting, and it'll pay you back in the details.

The inputs that matter

• prompt - the image to generate, max 800 chars. This is where the value of the tier shows: it follows complex, multi-clause instructions noticeably better than the standard model.
• width / height - 256 to 1536, step 8, default 1024×1024. Same ceiling as the standard Qwen tier; Max improves quality, not canvas size.
• seed - reproducibility, real. Use a fixed seed once you find a composition you like and want variations of.

Install

cd ComfyUI/custom_nodes
git clone https://github.com/eRepublik-Labs/comfyui-nodes-erpk.git erpk
cd erpk
pip install -r requirements.txt

Or ComfyUI Manager → search erpk → ERPK Custom Nodes. WaveSpeed API key with credits via right-click canvas > ERPK Settings (Settings > ERPK > API Keys).

Common issues

• "Prompt is required" - the usual free error for an empty box; the node validates before billing.
• No visible difference from the standard tier - Max's advantage shows in complex prompts. If your prompts are one-liners, you're paying for quality you're not exercising. Write fuller prompts or save the money.
• 1536px ceiling again - this is a Qwen API constraint across tiers. Need bigger? Seedream V4.5 in the same pack goes to 4096.
• Iteration cost - at $0.07/image, seed-locked variations add up. Batch your brainstorming on the $0.02 tier, then spend the Max credits on the shots that matter.

The mental model: standard Qwen Image for volume and drafts, Qwen Image Max for the hero images. It's the same folder in your workflow, just a different node and a bigger number on the bill.