Breaking

Z.ai Rolls Out New Flagship GLM-5.2 to Paying Plan Users

June 13, 2026 at 19:27 EDT

Foundation Models
Software Dev & Coding
Open Source

Z.ai (Zhipu AI) began immediate rollout of its new flagship AI model GLM-5.2 to all users of its GLM Coding Plans on June 13, 2026. The model emphasizes stronger coding performance, usable 1M-token context support and stability on long-horizon tasks, with open-sourcing and API access slated for the following week.

June 13, 2026 · Z.ai

GLM-5.2 Ships as Z.ai's New Open Flagship

China's first public AI company expands context to a 1M-token scale, sharpens coding and long-horizon agentic tasks, and reaffirms a fully open strategy — open weights under an MIT license expected next week.

usable context window (tokens)

5×

larger context vs prior 200K

MIT

expected open-weight license (~Jun 20)

Context Window: 200K → 1M Tokens

Same unit, drawn to scale — each block ≈ 200K tokens

200K

GLM-5

GLM-5.2

GLM-5 → GLM-5.2: What Changed

GLM-5 · Feb 2026

Context: 200K
744B total / 40B active MoE
Single thinking path
Open license

→

GLM-5.2 · Jun 2026

Context: 1M usable
MoE-based (details undisclosed)
Two modes: High / Max
MIT expected next week

Developer optimism

"Could rival Claude Opus" on coding
Open weights + low cost + long context seen as practical
Trained on domestic chips (Huawei Ascend)

Open questions

Detailed benchmarks not yet published
API access not available until next week
Weekend release leaves room for verification

The Stance

"Frontier intelligence must remain open-source — science should be global."

Released amid reports of U.S.-side restrictions on some frontier models, Z.ai commits to continued training on domestic Chinese chips and a fully open strategy under the MIT license.

Continue reading

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