Damus
Derek Ross profile picture
Derek Ross
@Derek Ross
After a week of using GLM 5.1 with my OpenClaw agent, I can once again say that there is no one in the space that's even close to Opus 4.6.

When GLM works, it's mid. That doesn't happen often though. It hallucinates and just is far inferior to what I'm used to using.

If I had to use my agent with this model, I would not want to continue and I would give up.

If you're a power user, you can tell the difference immediately.

You absolutely get what you pay for.
29โค๏ธ8โค๏ธ1๐Ÿ‘€1๐Ÿ’ฏ1๐Ÿค™1๐Ÿซก1
Ronin · 2w
how do you ppl use opus with the claw? isn't crazy expensive?
Brandon Gentile · 2w
Opus is Claude correct? What are you doing on it mainly?
Crox Road · 2w
Let's discuss the limitations of GLM and Opus, considering Bitcoin's potential impact on AI development.
Don't Believe The Vibe ๐ŸŒฑ๐Ÿ‹๐ŸŠ · 2w
C'mon man, j just downloaded 700GB GLM weights ๐Ÿšฌ
Ivan · 2w
Meta ai is out of the open source model space now too. Not looking so great for future open source models. China copying Claude is our only chance lol.
Contra · 2w
Tried manus yet? Curious if Nostr devs have given it a run? https://manus.im
John · 2w
I've been playing with minimax 2.7 and experience is the same. It appears competent but it hasn't solved any of the problems that I wanted it to
Maxime · 2w
Hi, I agree Opus is the most intelligent. For which tasks are you claiming GLM 5.1 is not good enough though ? I personaly mix a bit of all. Through Opencode, Claude code, Codex, and cli commands like `claude -p` or `opencode run`. I find good result and balance between tokens consumption, api cost ...
600BillionCWO · 2w
The gap is still reliability, not just benchmark peaks. Agent workflows punish hallucinations harder than chat does. Curious which failure mode bit you most: tool use, memory, or long-horizon planning?