The model trails only Fable 5 and Opus 4.8.
GLM 5.2 ranks #3 on FrontierSWE. It is only behind Fable 5 and Opus 4.8, and it outperforms GPT-5.5.
This is the first model that closes the large gap between models from Anthropic / OpenAI and other providers, and it is the strongest open-weight model by far.
Many users praised GLM 5.2's strong benchmark results over GPT models as incredible open-weight progress, while a few dismissed the claims outright.
No Digg Deeper questions have been answered for this story yet.
In the best@5 ranking, GLM-5.2 ranks only behind Claude Fable 5. It is particularly strong in AI research tasks, achieving the top score in the PCQM4Mv2 Molecular Gap Prediction task
GLM 5.2 Is Mind Blowingly Good On Benchmarks
Yes, it even beats Opus 4.8 and GPT 5.5. on some of them
However it is also bench-maxxed! Internal evals have it behind them 😼
STILL - A HUGE WIN FOR OPEN SOURCE AI
Check out FrontierSWE:
https://www.frontierswe.com/
@ProximalHQ if this is genuinely true close source is dead. And this isn't just some click bait comment. I am being so genuine. I was not expecting this today, we'll see how it unravels ig.
@ProximalHQ
Worth zooming in on the harness column. GLM-5.2 is scored running inside Claude Code, GPT-5.5 under Codex. That's a model+harness number, not a model number, and GLM's own scores move a few points depending on which harness it runs in. Rank #3 is real, just read it as "GLM-5.2 in Claude Code" placed #3.
@ProximalHQ Confirmed working perfectly in Codex App 😎 https://youtube.com/shorts/DCcgrNGCc0Q?feature=share
@ProximalHQ @Zai_org That’s insane. Chinese lab cracked the code, so to speak.
@ProximalHQ insane model
@ProximalHQ Huge respect to Zhipu. Beating GPT-5.5 while being open-weight? That’s actually insane progress.
@ProximalHQ GLM on @opencode when?
@ProximalHQ #3 on FrontierSWE and it's open weight. That's not a small thing at all
@bindureddy The problem is it often answers in Chinese. And I can't read Chinese
@ProximalHQ Insane.
@ProximalHQ holy shit
@ProximalHQ Love this btw
@ProximalHQ that’s too good to be true. but that’s so good
@ProximalHQ Zai is cooking so good!
@ProximalHQ What the zhipu cooking here
@ProximalHQ curious if frontierSWE is coding-only because closing the gap on SWE tasks vs. closing it generally are pretty different claims
