GLM-5.2

GLM-5.2 is positioned as a flagship model built for long-horizon tasks, representing a substantial capability jump over its predecessor GLM-5.1. Its defining feature is a solid 1M-token context designed to stably sustain extended workflows such as long, messy coding-agent trajectories, rather than simply accepting more tokens. The model is released under an MIT open-source license with no regional access restrictions, making the full weights and capabilities freely available to the research and developer community.

The architecture introduces IndexShare, a technique that reuses the same indexer across every four sparse attention layers, reducing per-token FLOPs by 2.9× at a 1M context length and making the long context practically efficient rather than merely theoretical. An improved MTP (multi-token prediction) layer increases speculative decoding acceptance length by up to 20%, further boosting inference throughput. GLM-5.2 also offers advanced coding capabilities with multiple thinking effort levels, letting users balance performance against latency depending on task complexity, making it well suited for sustained agentic coding workloads and other extended reasoning pipelines.

Ollama Cloudglm-5.2glm

Open provider page Provider docs

Quick Info

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

Cost

Input token cost: $1.40
Output token cost: $4.40

Limits

Output tokens: 131,072 tokens
Context window: 976,000 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

Ollama Cloud

Coverage

The U.S. National Institute of Standards and Technology's Center for AI Standards and Innovation (CAISI) published a public assessment of Z.ai's GLM-5.2 on July 17, 2026, evaluating the open-weight model released June 16, 2026 by the PRC-based developer (formerly Zhipu AI). CAISI found GLM-5.2 was probably the most cap The report details CAISI's independent evaluation methodology spanning overall capabilities, cyber capabilities, agent security, and safeguards, using item response theory across multiple benchmarks. CAISI notes that regardless of prompt-based safeguard robustness, safeguards for open-weight models can be circumvented

Ollama Cloud

CoverageBenchmark

Semgrep published a security research blog on June 22, 2026 reporting that GLM-5.2 outperformed Claude Opus 4.8 in their IDOR (Insecure Direct Object Reference) benchmark among open-weight models tested with identical prompts. The benchmark used the same dataset and prompt methodology Semgrep applies to frontier coding The evaluation compared GLM-5.2 against other popular open-source models in a controlled setup, with results that surprised Semgrep's security research team given GLM-5.2's open-weight status. The blog provides concrete technical evidence of GLM-5.2's agentic coding and vulnerability detection capabilities, complementi