AllImageAI

CogView4 AI Image Generator

Z.ai’s open text-to-image model for descriptive prompts, stylised scenes and classic diffusion tuning.

Developed by Z.aiActiveVerified August 25, 2026

Model Version

Mode

Preview mode: placeholder output, no credits charged

Output

CogView4

Z.ai’s open text-to-image model for descriptive prompts, stylised scenes and classic diffusion tuning.

SlowUp to 2K1:116:99:164:3

Try one of these prompts

Developer
Z.ai
Model
CogView4
Status
Active
Max Output
2K
Input
Text
Released
March 4, 2025
Last Verified
August 25, 2026

What Is CogView4?

CogView4 is Z.ai’s open text-to-image model for long descriptive prompts. The hosted operation exposes custom dimensions, negative prompts, CFG guidance, step count and deterministic seeds for traditional diffusion-style control.

What Is CogView4 Best For?

  • Long descriptive prompts
  • Stylised illustration
  • Controlled composition tests
  • Open-model workflows

How to Use CogView4

  1. Use a complete descriptive prompt.
  2. Add a concise negative prompt.
  3. Compare seeds at 1K.
  4. Render the selected composition at 2K if needed.

CogView4 Settings Explained

Steps

50 is the documented default; lower it only when faster drafts are more important than finish.

Negative Prompt

Remove unwanted objects, blur or watermarks without changing the main brief.

The panel above shows only the parameters CogView4 actually supports. Switch versions and unsupported controls disappear rather than sitting there doing nothing.

CogView4 Prompt Examples

Detailed vehicle scene

Cherry-red sports coupe parked on a coastal road at midday, polished body reflecting the blue sea, black racing rims, low camera angle, realistic commercial photography

Illustration

Layered paper-cut illustration of a mountain railway in winter, cobalt shadows, warm carriage windows, intricate handcrafted edges

CogView4 Test Results

Testing in progress

Every model runs the same eight-category prompt set — portrait, product, poster text, anime character, interior, multi-reference, image editing and character consistency — recording generation time, success rate, text accuracy and identity retention. That run has not finished for CogView4, so this page shows no scores. We do not publish numbers we have not measured.

CogView4 Limitations

  • Text-to-image only
  • The 50-step default is slower than distilled models
  • The hosted per-megapixel price makes 2K a final-only tier

CogView4 Pricing and Credits

Credits scale with resolution and image count. The Generate button always shows what this run costs.

1K

6 credits / image

2K

24 credits / image

Failed generations cost 0 credits.

Commercial Use and Image Rights

Commercial use is subject to the CogView model licence and the selected service terms.

  • Model used: CogView4
  • Model developer: Z.ai
  • The inference route is managed internally and may change without changing the selected model
  • Uploads are sent only to the configured inference service needed to complete the generation
  • You can delete any generation from your history at any time

Frequently Asked Questions

Who develops CogView4?+

The official repository is maintained by Z.ai and hosts the CogView model family.

Does a longer prompt help?+

The model documentation specifically recommends detailed prompts for better results.

Model Updates and Changelog

  • 2026-08-25 —— Added the exact CogView4 endpoint with full CFG, steps and negative-prompt mapping.