Don Codeleone
233 posts
@ollama Nothing says we distilled your model quite like this.
English
@guybedo @ivanfioravanti @Zai_org Are you using MAX thought? I'm wondering what the difference is between you and the other guy here. If you're both using something different, you're getting robbed.
English
this thing is a joke. Even when i was tokenmaxxing claude plan because they had reset the limits, i couldn't get more than 30% of the weekly in one day.
I subscribed to their plan yesterday and didn't event let the agents work during the night.
I'm at 73% of the quota.
That's just insane
English
I'm not facing GLM 5.2 Quota issues. I'm on the Max plan, but I've seen others reporting problems.
English
@SomeRandy11 @WoClaudecraft Well, at least you shared your honest opinion!
English
I tried @WoClaudecraft. An open-source MMO built with Fable 5.
And it already has the basic MMO loop:
▪️ quests, trading, rewards
▪️ duels
▪️ live multiplayer etc.
You can run around, interact with players and actually test the world.
Is it polished? No.
It stutters. Controls feel rough. Some parts feel like a tech demo.
But that’s the point.
One person built a playable MMO skeleton with AI.
Not a finished game. But enough to make the old rules feel outdated.
English
@Michel40715696 @OrganicGPT It's a copy-and-paste of American models; might as well just call it a bootleg, because that's what it is.
English
@SomeRandy11 @OrganicGPT GLM é um dos melhores modelos chineses atualmente, e seu uso é muita bom. Mas não é nem perto do Fable 5, mas chamar de lixo é muita ignorância
Português
From now on, I'll unfollow and block anyone who pushes for these nonsense benchmarks. GLM 5.2 > Fable 5, yeah, right 😏
Design Arena@Designarena
BREAKING: GLM-5.2 is now 1st on Design Arena. With an Elo of 1360, GLM-5.2 has jumped ahead of the now unavailable Claude Fable 5. And it's open weights. This is an improvement of 4 positions and 27 Elo points to achieve one of the highest Elo scores in our code categories since Design Arena started. Huge congratulations to the @Zai_org on the release!
English
@Nymne @Hesamation 1985 called, they want their Delorian back.
English
@SomeRandy11 @Hesamation I would say good coders don’t have to rely on a specific model.
If you are stuck with an expensive model because it mothers you, my condolences
English
GLM-5.2 doesn't need to beat Opus 4.8 when it’s 82% cheaper.
Prices for 1M output tokens:
> GLM-5.2: $4.40
> Claude Fable 5: $50
> Claude Opus 4.8: $25
> GPT-5.5: $30
Frontier AI is now luxury prices, and praying today's "subsidized" plans don't change to real prices tomorrow.
Lisan al Gaib@scaling01
GLM-5.2 genuinely looks like a good model and it's only 744B imagine what kind of margins the frontier labs have also try to imagine what a 1T Kimi or 1.6T DeepSeek model will do the only way for frontier labs to keep their lead is to scale their models and put in even more RL than these chinese labs
English
@ZixuanLi_ @ArtificialAnlys Hows that overpriced distillation model going? Better yet hows it feel selling a bootlegged American model 😂🗑️
English
Finally, Artificial Analysis Intelligence Index concludes the GLM-5.2 release.
Z.ai@Zai_org
Introducing GLM-5.2: Frontier Intelligence, Open Weights - Significant improvements in coding and agentic tasks - Strong long-horizon capabilities with a 1M context window - Two levels of reasoning effort: GLM-5.2 (max) pushes the limits, while GLM-5.2 (high) strikes a strong balance between performance and token efficiency - MIT-licensed open weights - Same API pricing as GLM-5.1 Tech Blog: z.ai/blog/glm-5.2 Weights: huggingface.co/zai-org/GLM-5.2 API: docs.z.ai/guides/llm/glm… Coding Plan: z.ai/subscribe Chat: chat.z.ai
English
@wamphotoai @xai Shut up and learn how to use it you fucken dinosaur.
English
@xai Hi,while I appreciate fast iterations, changing a model without the option of a fallback is not good. My results dramatically got worse (image to video) especially textures. I subscribed for SuperGrokHeavy for a certain quality ... that's now gone. If that does not change I' gone
English
Grok Imagine Video 1.5 is here
Our new image-to-video model with sharper realism, better physics and faster generations 🧵
grok.com/imagine
English
@Techmeme @jamesfraney Sounds like something a dictator does.
English
Sources: the Trump administration won't allow G7 countries to regain access to Mythos 5 as doing so would be "completely illogical"; UK requested a "carve-out" (@jamesfraney / New York Post)
(Visit Techmeme dot com for the link and full context!)
English
BREAKING: GLM-5.2 is now 1st on Design Arena.
With an Elo of 1360, GLM-5.2 has jumped ahead of the now unavailable Claude Fable 5.
And it's open weights.
This is an improvement of 4 positions and 27 Elo points to achieve one of the highest Elo scores in our code categories since Design Arena started.
Huge congratulations to the @Zai_org on the release!
English
@Officialtabsy @JasminLaine_ Holy, he's fat. Someone take the burgers from this man.
English
Prime Minister of Canada caught on a hot mic with President Trump.
He can clearly be heard saying: “less than 3% of our product”, “49,000 cars”, then while making a cut off gesture with his hands, “capped”, and “I thought you’d actually like that”, before walking away.
Presumably this is in relation to Chinese EV’s.
English
@JustKerp Please do add this to your Hermes queue. I know Im am, see you in 2 years.
English
@SomeRandy11 Sure man whatever you say...
Just don't start crying when you change your mind in 2 years and I pull these receipts
English
This new tech is about to change the industry forever
Credit - EmpereurDuSel on YT (Link Below)
#Hytale #HytaleModding
English
@Ananth7e @AiBattle_ It's just not worth using when you have Claude or GPT. Anything code-related, I wouldn't touch it.
English
@SomeRandy11 @AiBattle_ Yes, I do not wanna trust a lot these benchmarks. But GLM has generally been a good open-source model other than Kimi even for some serious tasks
English
@nirajxdev He paid for 2 years' worth of farmed datasets, not for the editor.
English
@Ananth7e @AiBattle_ No, GLM is just benchmaxing, it's nothing near 5.5 Xhigh in real-world coding tasks or image gen.
English