GLM 5.3 is the most impressive GLM model yet

GLM 5.3 is easily the most impressive GLM model yet. Excited to partner up and ship GLM 5.3 in Command Code AI soon.

A fun internal eval we run at Command Code: deliberately trap the model in a loop and see what happens. Every GLM model so far just... keeps looping. GLM 5.3 is the first one to notice, go “wait”, and break out.

It’s a small bench (15 loop traps), but escaping a loop you were intentionally placed in takes a little flicker of metacognition, and it’s fun to watch that show up.

Bonus: it plays very nicely with our new tool defer harness → ~50-60% token savings on an average session. Capability going up while cost going down is the good stuff.

GLM-5.3 at a glance

Frontier coding with emergent cyber capabilities.

Specs, pricing and plans are from the GLM-5.3 model page as of September 28, 2026, including any deal running that day, so they can differ from the launch pricing above. Intelligence and Coding Index scores are from Artificial Analysis.

Use GLM-5.3

Or type /model in a session and pick it. You can switch mid-session without losing context.

What real work costs on GLM-5.3

Try Command Code

The best coding agent for open models, in your terminal or on your desktop.

Desktop App · Docs · X · Discord