GLM-5.2 - The New Leading Open Weights LLM
Chinese AI lab Z.ai released GLM-5.2 to its coding plan subscribers on June 13th, then published the full open weights under an MIT license on June 16th. According to Artificial Analysis, one of the most widely respected independent benchmarks, GLM-5.2 is now the new leading open weights model on the Intelligence Index.
What Z.ai Shipped
GLM-5.2 is the same size as GLM-5.1 - 744B total parameters with 40B active (a Mixture of Experts design) - but it scores 11 points higher on the Intelligence Index v4.1. It is a text-only model with a 1M token context window, up from 200K on GLM-5.1, released under a permissive MIT license.
Where It Ranks
- Intelligence Index v4.1 - Scores 51, leading MiniMax-M3 (44), DeepSeek V4 Pro (44) and Kimi K2.6 (43).
- GDPval-AA v2 - Scores 1524 on this real-world agentic benchmark, ahead of MiniMax-M3 (1418) and DeepSeek V4 Pro (1328), and effectively level with GPT-5.5 (1514).
- Code Arena WebDev - Ranked 2nd for front-end web development, behind only Claude Fable 5 - impressive for a model with no image input.
Gains were strongest in scientific reasoning, including a 16-point jump on CritPt, a 12-point jump on HLE to 40 percent, and a 16-point jump on TerminalBench v2.1 to 78 percent.
The Tradeoff
GLM-5.2 is token-hungry. It uses about 43k output tokens per Intelligence Index task, up from 26k on GLM-5.1 and above MiniMax-M3 (24k) and Kimi K2.6 (35k). Even so, it sits on the Pareto frontier of intelligence versus cost, at roughly $0.46 per task - the lowest cost per task among models at its intelligence level. First-party API pricing is $1.4 / $0.26 / $4.4 per 1M input / cache hit / output tokens.
Run It on European Infrastructure
Because GLM-5.2 ships under an MIT license with open weights, you can self-host it instead of relying on a third-party API. Running it on your own hardware keeps prompts and data inside your jurisdiction - a real advantage for teams with GDPR or data-residency requirements. DCXV provides Tier III cloud servers in the EU built for exactly this kind of workload. Deploy a cloud server in Europe.
Bottom Line
GLM-5.2 is probably the most powerful text-only open weights LLM available today. It closes much of the gap to proprietary frontier models, ships under a permissive MIT license, and is available across providers including DeepInfra, Novita, Nebius, Fireworks and OpenRouter - or on your own infrastructure.
Source: Artificial Analysis, Simon Willison
