GLM-5.2

GLM-5.2 is presented by its publisher as a flagship model built specifically for long-horizon tasks, framed as a substantial leap over its predecessor GLM-5.1. The headline positioning centers on sustained quality across extended, messy coding-agent trajectories rather than simply accepting more tokens, and the model is marketed with a solid one-million-token context window that the team emphasizes remains stable under real agent workloads. Alongside the long-context push, GLM-5.2 introduces an advanced coding mode with multiple thinking-effort levels, letting users trade latency against reasoning depth depending on the task.

On the architecture side, Z.ai highlights an "IndexShare" design that reuses the same indexer across every four sparse attention layers, reportedly reducing per-token compute at long context, and an improved multi-token prediction layer aimed at speeding up speculative decoding. The model is released under an MIT license with weights published on Hugging Face and the codebase on GitHub, removing regional access limits and making it well suited for teams that need to self-host, fine-tune, or deeply inspect a coding-oriented agent model with a generous context budget.

Scalewayglm-5.2glm

Open provider page Provider docs

Quick Info

Provider: Scaleway
Model key: glm-5.2
Release date: Jun 13, 2026
Last updated: Jun 13, 2026
Input modalities
Output modalities
Capabilities

Cost

Input token cost: $1.80
Output token cost: $5.50

Limits

Output tokens: 16,384 tokens
Context window: 256,000 tokens

Transparent token rates

Compare GLM-5.2 pricing

Rates are shown per one million tokens. Combined means one million input plus one million output tokens.

Browse this family

Latest news about GLM-5.2

Scaleway

Official sourceAnnouncement

Scaleway's news post dated Jun 25, 2026 announces the immediate availability of GLM-5.2 on its Generative APIs, just 10 days after Z.ai released the model on June 16, 2026, and positions Scaleway as the first sovereign cloud provider in Europe to offer it. The article describes GLM-5.2 as built for the agentic era with The post claims GLM-5.2 currently ranks as the number one open-weight model for long-horizon coding tasks, citing a FrontierSWE result within 1% of Claude Opus 4.8 and ahead of GPT-5.5 and Gemini 3.1 Pro across major coding benchmarks. It also highlights the model's flexible thinking-effort levels and its MIT license w

Scaleway

Coverage

The U.S. National Institute of Standards and Technology's Center for AI Standards and Innovation (CAISI) published a public technical assessment of Z.ai's GLM-5.2 on July 8, 2026. The model was released as an open-weight release on June 16, 2026, and CAISI concluded that GLM-5.2 was probably the most capable open-weigh CAISI's assessment also flagged mixed safeguard behavior. The evaluation found that GLM-5.2's safeguards permit assistance with agentic cyber exploit development and block fewer sensitive biological questions than reference U.S. models, but that GLM-5.2 appears potentially more robust against agent hijacking and prompt