Breaking

Z.ai Releases GLM-5.2, Topping Open-Weight Models at 51 on Intelligence Index

June 17, 2026 at 03:13 EDT

Foundation Models
Open Source
Software Dev & Coding

Z.ai (formerly Zhipu AI) released its flagship LLM "GLM-5.2" in June 2026. It scored 51 on Artificial Analysis's Intelligence Index to lead open-weight models and ranked second on the Frontend Code Arena. The model ships as MIT-licensed open-weights, with the API available immediately (official announcement).

June 13, 2026 · Z.ai (Zhipu AI)

GLM-5.2 sets a new bar for open weights — first ever to break 50 on the Intelligence Index

The MIT-licensed flagship scores 51, lands around 4th overall, and ships a usable 1M-token context — pushing the open-weight camp deeper into Claude and GPT territory.

Intelligence Index — highest of any open-weight model

Usable context window (up from 200K)

744B

Total params · 40B active (MoE)

~1/10

Coding Plan cost vs Claude-class plans

Context window: a 5× leap

GLM-5.1 → GLM-5.2, measured in tokens

200K

GLM-5.1

GLM-5.2

Five times the headroom — credited with stabilizing large projects and long-horizon agentic workflows.

Benchmark snapshot

62.1

SWE-bench Pro (was 58.4)

82.7

Terminal Bench 2.1

91.2

GPQA Diamond

54.7

HLE with tools (40.5 base)

API price: $1.40 in / $4.40 out per 1M tokens · two thinking modes (High, Max).

What developers like

1M context stabilizes large projects & agent orchestration
Reported gains after migrating from Kimi K2.7 Code
Faster than the previous model

Open questions

Tight rate limits on cheaper plans; quota hit fast
Playable interactive games can fall short
Evals skew frontend — backend benchmarks wanted

From Vibe Coding to Agentic Engineering

A roughly two-month cadence — GLM-5 in February, GLM-5.1 in April, GLM-5.2 in June — leans on low pricing, open weights, and long context to pull developers away from costly closed models. The share race against Kimi K2.7, DeepSeek and MiniMax is just heating up.

Continue reading

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