GLM 5.2 scored positive where GLM 5.1 got 0%.
ok, here's how GLM 5.2 performs on a bench it definitely didn't see, and where GLM 5.1 scored 0.0%. Closer to Opus 4.8 than Sonnet 4.6 I hope their confidence seems more credible now
GLM 5.1 scores zero btw there's no way they benchmaxed this thing directly we shall see how 5.2 performs. I'd be surprised if it landed below MiniMax
Positive users praise GLM 5.2 for its strong unseen benchmark scores nearing Claude Opus and well-rounded capabilities like legal agentic tasks, while one negative reply mocks the high cost for only a tiny performance edge.
No Digg Deeper questions have been answered for this story yet.
Interesting that in "GLM 5.2 is on par with X model", X is so widely distributed over different capabilities, and it's not like the gap is just greater for the difficult ones. It plainly can't do some parlor tricks Sonnet 3.5 pulled off. In some hard model evals, it's around 4.8.
@teortaxesTex GLM 5.2 is such a crazy leap. MiMo V2.5 is the perfect pair for it to do the grunt work and vision while GLM 5.2 does the hard stuff
@teortaxesTex @yacineMTB they cooked hard sir. first chinese model that is actually well rounded and not only coding/math/agent focused?
@teortaxesTex The cost difference between GLM 5.2 & Opus 4.8 is insane for the minuscule performance gap on this benchmark lmao
@teortaxesTex Double gpt 5.5 ... Wow
@teortaxesTex link: https://www.vals.ai/benchmarks/hlab it appears to be a useful benchmark. matches my feelingsbench
@teortaxesTex @yacineMTB
@teortaxesTex Weird bench, why is gpt-5.5 so low
@teortaxesTex Very very interesting，thought it was only good at coding But also very good legal agentic capability
@teortaxesTex 这就叫含金量不仅看总分更得看戏份
@groundruled @teortaxesTex Game by GLM 5.2 Max(Deep Think) after a lot of iterations.
@teortaxesTex Its basically sonnet 4.8
@teortaxesTex glm 5.2 on a truly unseen bench is impressive if it holds. but "closer to opus" is a big claim - what's the actual score difference lookin like?
