GLM-Image
Zhipu AI · Undisclosed (MoE, full Ascend stack)
An open image model trained entirely on Ascend
An image-generation model open-sourced by Zhipu and Huawei in January 2026, trained end-to-end on Ascend: generation 94, Chinese 90, ecosystem 88 and a composite of 70.6. Its significance is not only the score — it is a sample of “domestic chips plus open weights” for image generation. Video 55 and agents 35 mark it as a creative tool.
Last updated Oct 21, 2026Editorial scoring
Bottom line
Eight-dimension breakdown
editorial mapping · 0–100- GLM-Image70.6
Weights: perception 20 / documents 18 / charts 14 / video 14 / agents 14 / generation 6 / Chinese 6 / ecosystem 8. Definitions and sources on the methodology page.
Public benchmarks
as reported, not converted| Benchmark / metric | Value | Note |
|---|---|---|
| 训练芯片 | 昇腾 | End-to-end |
| 许可证 | 社区许可 | Commercial use, not OSI |
| 上下文 | 32K | For generation tasks |
Public scores come from different versions and prompts and are not interchangeable with our composite; they are one evidence input.
Scenario mapping
five stress tests, one scale| Scenario | Score | Observation |
|---|---|---|
| Chinese poster generation | 94 · Good layout and text | |
| Product-image generation | 90 · Stable materials and lighting | |
| Chinese receipt OCR | 78 · Not an OCR model | |
| 30-min video QA | 55 · Not a video model | |
| Chinese cultural context | 90 · Good Chinese-style elements |
Strengths and limitations
Strengths
- Full Ascend training plus open weights is a scarce domestic-chip image-generation sample.
- Generation 94 and Chinese 90 work for Chinese posters and product images.
- Ecosystem 88: smooth integration with Zhipu tooling.
Limitations
- Average understanding: documents 78, charts 65.
- Agents 35 and video 55: no business automation.
- The Zhipu community licence is not OSI open source; check before commercial use.
Price and real cost
List price
—N/A
Open weights; self-hosted, no API fee. For 1,000 images at 1,024px, ~1,100 image tokens each and 300 output tokens, the monthly cost is about ¥0.0 (assuming input at 25% of output price).
Use your volumeAbout GLM-Image
Is GLM-Image open source?
Licence: Open weights (Zhipu community licence). Weights are available, but check the exact terms and filing requirements before commercial use.
How much does GLM-Image cost?
Open weights; self-hosted, no API fee. It ranks #51 with a composite of 70.6. Real bills depend on resolution, tiling, caching and volume discounts.
What is GLM-Image best for?
Image generation in domestic-chip environments; Chinese posters and product images; Creative tooling inside the Ascend ecosystem。
Why does GLM-Image score 70.6?
The composite weights eight dimensions: perception 20%, documents 18%, charts 14%, video 14%, agents 14%, generation 6%, Chinese 6%, ecosystem 8%. Its strongest dimension is Generation & editing; its weakest is GUI agents & tool use.
Personal ranking · Not an official score
Every ranking, score and comparison on this page is an independent personal ranking, not an official score, and represents no vendor, institution or regulator. Scores are a subjective mapping of public evidence under our own weights — change the weights and the order changes. Disagree, or spotted a factual error? Contact the site owner: jizhi2024@foxmail.com — we check each report and log corrections in the changelog.