The joke claims fictional GLM writes cleaner, minimalist code.
glm5.2 (max) is better than gpt5.5 (xhigh) at coding. it has better taste, I don't have to constantly redirect it away from nonsensically defensive slop coding the way i do with gpt5.5. it's much better at clean, minimalist code that understands the spirit of what i'm going for.
Positive users praise GLM 5.2 for outperforming GPT 5.5 at coding with better taste and advancing open models, while negative users criticize its slowness and limited subscription access.
No Digg Deeper questions have been answered for this story yet.
Oh shit
glm5.2 (max) is better than gpt5.5 (xhigh) at coding. it has better taste, I don't have to constantly redirect it away from nonsensically defensive slop coding the way i do with gpt5.5. it's much better at clean, minimalist code that understands the spirit of what i'm going for.
@matthewdif It’s a good manager for 5.5 though bc it does make better decisions
@FullOfCaffeine @matthewdif I use http://z.ai directly. Here is the setup using opencode
@matthewdif What's the best provider for GLM 5.2 at the moment, in your opinion/knowledge?
@matthewdif GLM 5.2 is very robust I’m like it more the more I use it
Increadible that this is an open weights models
Zai did something incredible. Im excited for the future
@yacineMTB Remember, local models (even open source to an extent) are less at risk for context manipulation nonsense (mainly thinking pre-prompts or injections). Having pristine context to work with allows more focused inference.
@matthewdif What harness are you using? OpenCode?
@TrickRiggin yes
@mov_axbx atp just get b300s
@matthewdif but on Max, GLM thinks 2x to 3x slower than GPT-5.5 xhigh. Are you okay with that?
@yacineMTB
@matthewdif I find that it catches major bugs that opus 4.8 max misses as well.
@matthewdif Subscriptions are very difficult to get. It seems that their computing resources are limited.
@matthewdif I felt that when you said defensive coding, gotta try glm
@matthewdif Too bad it’s not multimodal
@matthewdif It's quite slow though..i kept on checking to see where it's going and the thing is very slow.
Quality wise not sure haven't used GLM5.2 enough..but it's one of the better ones for sure.
@yacineMTB Looks like I’m going to have it at Q5, 1M ctx and nobody can take it away from me
You thought GPUs were $$$ now, wait until people catch on. If we get a <=1T model at Fable ability, RTX Pros are going to $25k.
@matthewdif how about the non max version, can you detect a difference?
@alexbastian_ai it's not a problem for how i am using it atm, while it thinks for the next step i spend time reviewing all the code it just wrote. hopefully the providers will get faster over time though.
Yes. Nothing more irritating than watching gpt 5.5 normalize the same string 15 different times.
I like glm 5.2 also but just Pro tip, gpt 5.5 rarely does better beyond medium. Extra high is just likely to make it over complicate or go off on tangents. It's like as an obligation to think more about stuff so it puts more nonsense in that could never actually happen.
