Breaking

GLM-5.2 Lands 5th in Design Arena Mobile App Dev at Elo 1257

June 23, 2026 at 18:38 EDT

Foundation Models
Software Dev & Coding
Open Source

GLM-5.2, the open-weight model released by Z.ai, has reached Elo 1257 and 5th place in Design Arena's "Mobile App Dev" category. The score marks a two-position jump from the prior GLM-5.1 and puts the model in the same performance band as Anthropic's Claude Sonnet 4.6 and Opus 4.6.

June 2026 · Z.ai · GLM-5.2

Open Weights Reach the Design Frontier

GLM-5.2 jumps to 5th in Design Arena's Mobile App Dev category with an Elo of 1257 — landing in the same band as Claude Sonnet 4.6 and Opus 4.6, and making Z.ai the top open-weight lab in the category.

5th

Mobile App Dev rank (up from 7th)

1257

Elo, Mobile App Dev category

1360

Elo, Design Arena overall (top rank)

Context window: ~5× larger

Practical context expands from ~200K to ~1M tokens — enabling stable repository-scale work.

GLM-5.1

~200K

1 block

GLM-5.2

~1M

5 blocks

Architecture

744B total / ~40B active

Mixture-of-Experts · 131,072-token output cap · two-level thinking effort

Price & License

~$4.40 / 1M out

MIT open weights · roughly one-fifth the cost of closed models

Where it shines

Agentic coding matching or surpassing Claude Opus
Stable repository-scale work via 1M-token context
Beat Claude on a cyber-defense benchmark in testing
Runs locally on a single 256GB Mac (quantized)

The caveats

Some tasks need extra guidance or multiple review passes
A tendency to mix in Chinese has been reported
Benchmarks weigh human preference over pure code correctness

Continue reading

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