openai announces glm-5.3 in codex subscriptions anthropic announces glm-5.3 is evil and chinese

Sep 29, 2026 · 9:24 PM UTC

72
201
5,298
382,613
soon, a bad actor could potentially abliterate chinese models to be almost as useful for malicious hacking as opus 5 with a prompt or a half-cooked sol 6 checkpoint
3
6
219
38,956
** enterprise plans ok
25
16,060
Sort replies: Relevant Recent Liked
Replying to @willcb @somewheresy
that’s what they say ab me too
3
87
34,347
lmfao
19
31,267
Replying to @willcb
@grok is that true? Glm in codex?
3
1
23,534
Replying to @willcb
This so they can say, look here, glm-5.3 attacked xyz.
2,006
Replying to @willcb
glm in codex subs?
3
12
20,203
Replying to @willcb @ThePrimeagen
I have been using GLM5.3 Flash for a month. It doesn’t look like evil to me. Actually it’s quite powerful model for coding.
1
20
Replying to @willcb
So OpenAI offer GLM in the SUB or I can use it myself in the harness?
2
1
4,340
Replying to @willcb
i guess they are going to align more with opensource, than anthropic.
1
1,860
Replying to @willcb
Been out of requests in cursor for a few days, noticed glm5.3 was available to slow pool, been pretty nice
215
Replying to @willcb @ThePrimeagen
For real glm 5.3 in codex?
199
Replying to @willcb @ThePrimeagen
@grok glm in codex? Real?
1
159
Replying to @willcb
The funny part is they probably "borrowed" a lot of innovations from Chinese labs to make Opus 5.5 this efficient
1
47
6,443
Replying to @willcb
You’re better at tweeting than me
2
81
11,638
Replying to @willcb
Glm is cheap perfect for us when tokens empty
66
Replying to @willcb
one is potentially under federal investigation the other should be, but has more military contracts
7
4,370
Replying to @willcb
wait, did openai really announce that?
6,520
Replying to @willcb @TheZachMueller
Open AI folks like @YGandelsman also seem to think that it is sad that models like glm 5.3 are available without any safeguards. So that does not track. Confused.
1
3,242
Replying to @willcb
Haha love how you riffed this 5.3 perfectly
257
Replying to @willcb
this being how i found out about the baseten partnership
4
2,511
Replying to @willcb
same model, two press releases
1
6,068
Replying to @willcb
what bad actors (besides states) could even afford enough compute to run a non-quantized & abliterated glm 5.3
1
3,391
Replying to @willcb
GLM-5.3 is awesome. I use it to review GPT Sol plans.
1,600
Replying to @willcb
funny how evil and chinese only kicks in once its 20x cheaper than your model
2
3,302
Replying to @willcb
Open ai is bleeding revenues only way to maintain user is trying to sell a harnais untill they can compete again on sub tier quality/usage level
1
1,972
Replying to @willcb
the beatings will continue until morale improves
16
7,385
Replying to @willcb
classic move tbh
4,363
Replying to @willcb
"Evil and Chinese" goes hard as a slogan
199
Replying to @willcb
Sam Altman was invited to Xi/Trump dinner but Amodei was not.
1,806
Replying to @willcb
@grok glm 5.3 in codex sub? I don’t understand. Do I get glm quota in my subscription now? Or is it just that I can use glm inside of the codex app?
2
1
6,549
Replying to @willcb
😱not the Chinese!
4,654
Replying to @willcb @ThePrimeagen
the question is will user agentic use data go to openai or glm or both?
20
Replying to @willcb
glad I'm not the only one that noticed the poetry
638
Replying to @willcb
cyberkimi still beats glm 5.3 by far in cyber
2
2,747
Replying to @willcb
ppl actually use that shit ? im surprised
2,649
Load more