Top AI Product

Z.ai (Zhipu AI) released GLM-5.2 on June 13, and the story is the gap that isn’t there: on FrontierSWE it scores 74.4% against Claude Opus 4.8’s 75.1% — a near-tie with the proprietary leader, from an open-weight model you can download.

## The benchmarks

GLM-5.2 is a 753-billion-parameter model under an MIT license with a 1M-token context window, tuned coding-first. On SWE-bench Pro it hits 62.1 to Opus 4.8’s 69.2; on Terminal-Bench 2.1 it lands at 81.0, within four points of Opus. It also matches or beats GPT-5.5 on several long-horizon coding benchmarks — at roughly one sixth of the cost.

## API and access

It ships through the GLM Coding Plan with open weights, and a standalone API rolled out in the days after launch, so teams can self-host or call it directly. The MIT license is the aggressive part: no usage gates, no weight restrictions. The catch worth naming is data — the hosted API routes your code to a China-based provider, which is exactly why the downloadable weights matter most to anyone who can run them.