Term
GLM-5.2
GLM-5.2 (June 2026) is Zhipu AIs (Z.ai) open-weight model — an MoE with around 744B parameters (~40B active), a 1M-token context and MIT license, strong at agentic coding.
GLM-5.2 — explained in more detail
GLM-5.2 is an open-weight language model from the Chinese lab Zhipu AI (operating internationally as Z.ai), released in mid-June 2026. The GLM line (General Language Model) is one of the established open model families from China; 5.2 is an evolution of GLM-5.1. The model is under the permissive MIT license and downloadable with full weights via Hugging Face — so it can be run locally or on your own hardware, unlike pure API models.
Architecturally, GLM-5.2 is a Mixture-of-Experts (MoE) with around 744 billion total parameters, of which only about 40 billion are active per token — keeping inference cost manageable despite the large capacity. The context window is around 1 million tokens. Its clear focus is agentic coding: in independent tests GLM-5.2 reached, among others, 81.0 on Terminal-Bench 2.1 and 62.1 on SWE-bench Pro, beating GPT-5.5 on several long-horizon coding benchmarks at about one sixth of the cost. On the Artificial Analysis Intelligence Index it scored 51, making it the strongest open-weight model of its time. As a hosted service the price is around 1.40 US dollars per million input and 4.40 US dollars per million output tokens.
Example / Practical context
GLM-5.2 is mainly attractive for teams that want near-frontier coding performance without locking into a closed API provider. A typical pattern: a dev team runs the model itself (or via a cheap host) to power a coding agent working across a large repository — reading files, making changes, running tests and iterating. The long context holds the relevant code, and the open license allows operation with no data flowing to third parties. Because of its strong price-performance ratio, GLM-5.2 is frequently cited as a cheap alternative to GPT or Claude models for coding workloads.
Distinction from related terms
Within the GLM line, 5.2 follows GLM-5.1 and GLM-4.6 and mainly raises the coding and agent capabilities. The central difference from API-only models (GPT, Claude, Gemini, Qwen-Max) is the open weights under the MIT license — self-hosting is possible. Among open frontier models, GLM-5.2 competes with DeepSeek, Qwen (the open line), Kimi (Moonshot) and MiniMax; its profile is the combination of strong coding performance, long context and low cost. Open weight is not the same as fully open source, though: training data and code usually remain undisclosed.