Breaking

Z.ai's GLM-5.2 Lands 5th in Design Arena's Mobile App Category

June 26, 2026 at 14:36 EDT

Foundation Models
Open Source
Software Dev & Coding

GLM-5.2, the open-weight large language model released by China's Z.ai, has reached 5th place in the Mobile App category of the crowdsourced evaluation site Design Arena, posting an Elo rating of 1248. The result marks a two-position jump from the previous GLM-5.1 and places the model in the same performance band as Anthropic's Claude Sonnet 4.6. In this category, Z.ai ranks as the top open-weight lab and third overall.

June 2026 · Z.ai (Zhipu AI)

GLM-5.2: Open Weights Reach the Design Frontier

Z.ai's MIT-licensed 744B MoE model climbs to 5th in Design Arena's Mobile App leaderboard — two spots above GLM-5.1 and into the same band as Claude Sonnet 4.6 — while reportedly matching closed frontier models at a fraction of the cost.

1248

Mobile App Arena Elo — 5th place

744B

MoE parameters (~40B active)

Token context window

~1/6

Reported cost vs GPT-5.5 on long-horizon coding

Terminal-Bench 2.1 — score to scale

Higher is better. Each block = ~10 points.

62.0

GLM-5.1

81.0

GLM-5.2

85.0

Opus 4.8

SWE-bench Pro also rises to 62.1 (from GLM-5.1's 58.4), closing the gap with closed frontier models.

What developers praise

"First open-weight model that can replace Opus 4.8 or GPT-5.5 for daily use"
Handles whole repos within 1M context; idiomatic code
Cost-efficient — ~$3.36 for 6M tokens cited
Strong for UI redesign, data analysis, bug hunting

Noted limitations

Slow and verbose; over-plans on some tasks
Higher token use under Max effort
No vision support — can't read screenshots
Full local run impractical; drift on long tasks

Where it stands

Top open-weight lab in Design Arena — and 3rd overall, behind only Google DeepMind and Anthropic.

API pricing held at GLM-5.1 levels:

$1.40 / $4.40 per M tokens (in / out)

Web Dev & Full-Stack rank ~Elo 1288 · MIT open-weight license

Continue reading

The rest of this article is for AI News Blitz readers. Choose an option below to keep reading.