Top
Best
New

Posted by pella 19 hours ago

GLM-5.3: Frontier coding with emergent cyber capabilities(z.ai)
1020 points | 502 commentspage 7
petesergeant 15 hours ago|
Their own hardness (ZCode) seems to be a GUI, which doesn't work for me. They say they support other harnesses. However, it seems like I can inject the plan into other harnesses, like Claude Code[0]. Does anyone who's been using GLM models for a while have a strong feeling for if it does better in some harnesses than others, or should I just use my favourite harness?

0: https://docs.z.ai/devpack/tool/others

ljosifov 14 hours ago||
I've used GLM-s the longest with Claude Code and their Anthropic supplied endpoint. As per their docs

$ ANTHROPIC_BASE_URL="https://api.z.ai/api/anthropic" ANTHROPIC_AUTH_TOKEN="zai-api-key" claude --dangerously-skip-permissions

Lately I use Zai in omp (oh-my-pi). It's listed built-in provider can be selected without configs shenanigans. Fits in the overall setup e.g. can select GLM-5.2 (now 5.3), and assign it role [plan] or [advisor]. I got reminded now of glm-5v-turbo. Think that 'v' was for vision. Assigned it role [vision] in omp now, let's see what happens. :-)

surgical_fire 14 hours ago|||
I am using GLM on Pi without any issues. You just create an API key.

Started recently though, mostly been using GLM 5.2 for planning with DeepSeek V4-flash for implementation.

scotty79 15 hours ago||
I use it with random harnesses. It behaves consistently.
bsenftner 13 hours ago||
So, "cyber capabilities", whoa there horsey, what the fuck is that? Are we making up words or are you trying to court the black hat crowd?
danggggg 13 hours ago|
[dead]
petesergeant 15 hours ago||
[total rewrite: their subscription code is buggy. It takes a while for paid subscriptions to show up, and for upgrades to take effect. Original comment was whining about this]
unrvl22 15 hours ago|
their sub is crap. use opencode go (multiple workspaces) or wait for weights to drop
MrBuddyCasino 18 hours ago||
An I the only one who was disappointed with GLM 5.2 after all the hype? It was thinking forever and sometime just stopped mid task.
indigodaddy 8 hours ago|
That's mostly about 1) harness incompatibility with the GLM and/or 2) Bad implementation of hosting the model by your upstream vendor
MrBuddyCasino 9 minutes ago||
I was using it via OpenCode and OpenRouter. What setup do you use?
steffi_oliver 7 hours ago||
[dead]
Satoshi_Bro 11 hours ago||
[flagged]
jocelyner 17 hours ago||
[dead]
libertas_quae_s 15 hours ago||
[dead]
zaj00l 9 hours ago||
[dead]
bigwheel 13 hours ago|
[dead]
More comments...