It would be nice if you can test the model with different harnesses, Z.ai's own Z Code, Claude Code, Open Code, Pi, Cursor etc.

My impression is that the choice of harness matters a lot.

I'e been using their models pretty much daily for the past 2 months to work on the codebase of a very complex B2B2C platform written in an unusual functional language (F#) with an angular frontend.

I also use Claude premium daily for another client, and i use Codex. and i can tell you that GLM5 is at this point much more capable than Claude and Codex for complex backend end work, complex feature planning, and long horizon tasks. One thing i've noticed is that it is particularly good at following instructions and guidelines, even deep into the execution of a plan.

To me the only problem is that z.ai have had trouble with inference : the performance of their API has been pretty poor at times. It looks like this is an hardware issue related to the Huawei chips they use rather than an issue with the model itself. The situation has been substantially improving over the past few weeks.

GLM5.1, GLM5-Turbo and GLM5v are at this point better than Opus, Codex, Gemini and other claude source models. We have reached a major turning point. To me, the only closed source model still in the game is codex as it is much faster at executing simple tasks and implementing already created plans.

Try GLM5v for your PDF work, it's their last generation vision model that has been released a couple of days ago.

From what I gather qwen is currently the undisputed local LLM king.

I still use GLM 4.7 for well defined coding tasks. I never got 5.0 to work satisfactorily, it felt like a hosting problem (z.ai) where it would work for a while then, for whatever reason, it couldn't respond to the context any more - but that's just a hunch.

I had no such trouble with 4.7 and find it fast and productive. Haven't tried 5.1; am using openAI models for coding most of the time.

I don't agree. I think their models are pretty good. The company's infrastructure though seems to be so so.

I do not know if it is good, because I have not tested it yet, but the most recent uncensored model is:

https://huggingface.co/trohrbaugh/gemma-4-31b-it-heretic-ara...

which was produced immediately after Google released their new Gemma 4 model.

>by the way, which is the best uncensored model at the moment?

There are no such models, depending on your definition of censorship. If you're referring to abliteration and similar automated techniques, they're snake oil.

Completely agree with this statement "Z.ai and their GLM models are pretty low quality." I have been trying out and it's kind of useless compare to SOTA models.

Surely at this point it’s part of the training set and the benchmark has lost its value?

It made it realistic. A pelican is much more likely to be flying in the sky than riding a bicycle.

Simon, you need to come up with improved benchmarks soon.

SSD offload is always a possibility with good software support. Of course you might easily object that the model would not be "running" then, more like crawling. Still you'd be able to execute it locally and get it to respond after some time.

Meanwhile we're even seeing emerging 'engram' and 'inner-layer embedding parameters' techniques where the possibility of SSD offload is planned for in advance when developing the architecture.

This has got to be bait..

1) OpenAI and Anthropic are killing it, and continue to do so, their coding tools are unmatched for professionals.

2) Local models don't hold a candle to SOTA models and there's nothing on the horizon that indicates that consumers will be able to run anything close to what you can get in a data center.

3) Coding is a killer product, OpenAI and Anthropic are raking in the cash. The top 3 apps are apps in the app store are AI. Everyone who knows anything is using AI, every day, across the economy.

No killer product? Coding assistants and LLM's in general are the single most awe-inspiring achievement of humanity in my lifetime, technological or otherwise. They've already massively improved my and others' lives and they're only going to get better. If pre and post industrial revolution used to be the major binary delineation of our history, I'm fairly confident it will soon be seen as pre and post AI instead.

I don't see how its possible to think this. AI coding assistants are some of the most useful technologies ever created, and model quality is by far the most important thing, so I doesn't make sense why local inference would be the path forward unless something fundamentally changes about hardware.

> no moat

I'd like to think the superior product wins. But Windows still thrives despite widespread Linux availability. I think sometimes we can underestimate the resilience of the tech oligopolies, particularly when they're VC-funded.

Google doesn't release Gemma 4 if Gemini is similiar good.

We probably talk abuot a year of progress diffeerence.

Its also still quite expensive for an avg person to consume any of it. Either due to hardware invest, energy cost or API cost.

Also professionally I don't think anyone will really spend a little bit less money of having the 3th quality model running if they can run the best model.

I'm happy that we reach levels were this becomes an alternative if you value open and control though.

What benefit is there to dropping $50k on GPUs to run this personally besides being a cool enthusiast project?

(1) is absolutely not true if you actually use these models on a regular basis and include Google in here too. The difference in reliability beyond basic tasks is night and day. Their reward function is just so much better, and there are many nuanced reasons for this.

(2) is probably true but with caveats. Top-tier models will never run on desktop machines, but companies should (and do) host their own models. The future is open-weight though, that much is for sure.

(3) This is so ignorant that others have already responded to it. Look outside of your own bubble, please.

I was trying to use Claude.ai today to learn how to do hexagonal geometry.

Every time I asked a question it generated an interactive geometry graph on the fly in Javascript. Sometimes it spent minutes compiling and testing code on the server so it could make sure it was correct. I was really impressed.

Anyway I couldn't really learn anything since when the code didn't work I wasn't sure if I had ported it wrong or the AI did it wrong, so I ended up learning how to calculate SDF and pixel to hex grid from tutorials I found on google instead.

Posted this after mythos came out? The hutzpah

No moat: yes. Cooked: no. It's a race. Why assume they're going to lose? It relies on (2) which is only true if AI usefulness plateaus at some level of compute. That's a huge claim to be making at this stage. (3) AI has lots of killer products already. The big one is filling in moats. Unrealized potential though for sure.

>(1) OpenAI & Anthropic are absolutely cooked; it's obvious they have no moat

I think big corporations will continue to use them no matter how cheap and good other models are. There's a saying: nobody was fired for buying IBM.

How good would open source models be if they couldn't distill higher quality private models?

(3) is simply a lie spread by engineers who have no other context. I manage some real estate (mid-term rentals) and everyone I know has switched over to AI robo-handlers to do the contact at this point. It's almost a passive investment at this point. Some can even handle interfacing with contractors and service requests for you. Revolutionized the field in my opinion.

The model is the killer product

I just set the context window to 100k and manage it actively (e.g. I compact it regularly or make it write out documentation of its current state and start a new session).

For me, Opus 4.6 isn't working quite right currently, and I often use GLM 5.1 instead. I'd prefer to use peak Opus over GLM 5.1, but GLM 5.1 is an adequate fallback. It's incredible how good open-weight models have gotten.

When it works and its not slow it can impress. Like yesterday it solved something that kimi k2.5 could not. and kimi was best open source model for me. But it still slow sometimes. I have z.ai and kimi subscription when i run out of tokens for claude (max) and codex(plus).

i have a feeling its nearing opus 4.5 level if they could fix it getting crazy after like 100k tokens.

Why is that sad? A free and open source model outperforming their closed source counterparts is always a win for the users

After the context gets to 100k tokens you should open a new session or run /compact.

I honestly still hold onto habits from earlier days of Claude & Codex usage and tend to wipe / compact my context frequently. I don't trust the era of big giant contexts, frankly, even on the frontier models.

I've set max context to 180k and usually compact around 120k. It's much better to re-read stuff than to have it under-perform when it's over 120k.

Isn't the same with opus nowadays?

Yep, haven't tried 5.1 but for my PHP coding, GLM-5 is 99% the same as Sonnet/Opus/GPT-5 levels. It is unbelievably strong for what it costs, not to mention you can run it locally.

I am working on a large scale dataset for producing agent traces for Python <> cython conversion with tooling, and it is second only to gemini pro 3.1 in acceptance rates (16% vs 26%).

Mid-sized models like gpt-oss minimax and qwen3.5 122b are around 6%, and gemma4 31b around 7% (but much slower).

I haven’t tried Opus or ChatGPT due to high costs on openrouter for this application.

It really bothers me that people refer to open weight models as being open source. They fundamentally aren't and are more akin to freeware than anything else.

Same thing I noticed.

My use cases are not code editing or authoring related, but when it comes to understanding a codebase and it's docs to help stakeholders write tasks or understand systems it has always outperformed american models at roughly half the price.

Blog post is new but the model is about 2 weeks in public.

My local tennis court's reservation website was broken and I couldn't cancel a reservation, and I asked GLM-5.1 if it can figure out the API. Five minutes later, I check and it had found a /cancel.php URL that accepted an ID but the ID wasn't exposed anywhere, so it found and was exploiting a blind SQL injection vulnerability to find my reservation ID.

Overeager, but I was really really impressed.

It's been out for a while.

The (none) version especially shows considerable degradation.

I'm on their pro plan and I respectfully disagree - it's genuinely excellent with GLM 5.1 so long as you remember to /compact once it hits around 100k tokens. At that point it's pretty much broken and entirely unusable, but if you keep context under about 100k it's genuinely on par with Opus for me, and in some ways it's arguably better.

I am on the mid tier Coding plan to trying it out for the sake of curiosity.

During off peak hour a simple 3 line CSS change took over 50 minutes and it routinely times out mid-tool and leaves dangling XML and tool uses everywhere, overwriting files badly or patching duplicate lines into files

My impression is that different users get vastly different service, possibly based on location. I live in Western Europe, and it works perfectly for me. Never had a single timeout or noticeable quality degradation. My brother lives in East Asia, and it's unusable for him. Some days, it just literally does not work, no API calls are successful. Other days, it's slow or seems dumber than it should be.

Every model seems that way, going back to even GPT 3 and 4, the company comes out with a very impressive model that then regresses over a few months as the company tries to rein in inference costs through quantization and other methods.

This is surprising to me. Maybe because I'm on Pro, and not Lite. I signed up last week and managed to get a ton of good work done with 5.1. I think I did run into the odd quantization quirk, but overall: $30 well spent

I'm also on the lite plan and have been using 5.1 for a few days now. It works fine for me.

But it's all casual side projects.

Edit: I often to /compact at around 100 000 token or switch to a new session. Maybe that is why.

I'm on their lite plan as well and I've been using it for my OpenClaw. It had some issues but it also one-shotted a very impressive dashboard for my Twitter bookmarks.

For the price this is a pretty damn impressive model.

Is there any advantage to their fixed payment plans at all vs just using this model via 3rd party providers via openrouter, given how relatively cheap they tend to be on a per-token basis?

Providers like DeepInfra are already giving access to 5.1 https://deepinfra.com/zai-org/GLM-5.1

$1.40 in $4.40 out $0.26 cached

/ 1M tokens

That's more expensive than other models, but not terrible, and will go down over time, and is far far cheaper than Opus or Sonnet or GPT.

I haven't had any bad luck with DeepInfra in particular with quantization or rate limiting. But I've only heard bad things about people who used z.ai directly.

> Obvious quantization issues

Devil's advocate: why shouldn't they do it if OpenAI, Anthropic and Google get away with playing this game?

I have their most expensive plan and it's on-par and sometimes better than Claude although you have to keep context short. That being said, the quota is no longer generous. It's still priced below Claude but not by that much. (compared to a few months ago where your money gets you x10 in tokens)

I'm on their Lite plan and I see some of this too. It is also slow. I use it as a backup.

It has been useless for long time when compared to Opus or even something like Kimi. The saving grace was that it was dirt cheap but that doesn't matter if it can't do what I want even after many repeated tries and trying to push it to a correct solution.

There is also GLM-5-Turbo, have you tried it for your claw?

Gemini and GPT have Deep Research models already, Mythos looks like much the same thing.

In my opinion it would be way cooler if it actually created a real Linux desktop environment instead of only a replica.

Would it succeed? Probably not, but it would be way more interesting, even if it didn't work.

I find things like Claude's C compiler way more interesting where, even though CCC is objectively bad (code is messy, generates very bad unoptimized code, etc) it at least is something cool and shows that with some human guideance it could generate something even better.

It is a big claim without the source and prompting.

Mythos is most definitely not in response to this announcement.

I've done some very long sessions on OpenCode with Dynamic Context Pruning. Highly recommend it.

https://github.com/Opencode-DCP/opencode-dynamic-context-pru...

> It's an okay model. My biggest issue using GLM 5.1 in OpenCode is that it loses coherency over longer contexts

Since the entire purpose, focus and motivation of this model seems to have been "coherency over longer contexts", doesn't that issue makes it not an OK model? It's bad at the thing it's supposed to be good at, no?

That's pretty few, at least for the way I'm currently using LLMs. I have them do some Nix work (both debugging and coding) where accuracy and quality matters to me, so they're instructed to behave as I would when it comes to docs, always consulting certain docs and source code in a specific order. It's not unusual for them to chew through 200k - 600k tokens in a single session before they solve everything I want them to. That's what I currently think of when I think of "long horizon within a single context window".

So I need them to not only not devolve into gibberish, but remain smart enough to be useful at contexts several times longer than that.

Yes, this is frustrating, but it doesn’t occur in CC. I run the conversation logs through an agent and opencode source, and it identified an issue in the reasoning implementation of opencode for Zai models. Consequently, I ceased my research and opted to use CC instead.

Chiming in to second this issue. It is wildly frustrating.

I suspect that this isn't the model, but something that z.ai is doing with hosting it. At launch I was related to find glm-5.1 was stable even as the context window filled all the way up (~200k). Where-as glm-5, while it could still talk and think, but had forgotten the finer points of tool use to the point where it was making grevious errors as it went (burning gobs of tokens to fix duplicate code problems).

However, real brutal changes happened sometimes in the last two or three months: the parent problem emerged and emerged hard, out of nowhere. Worse, for me, it seemed to be around 60k context windows, which was heinous: I was honestly a bit despondent that my z.ai subscription had become so effectively useless. That I could only work on small problems.

Thankfully the coherency barrier raised signficiantly around three weeks go. It now seems to lose its mind and emits chaotic non-sentance gibberish around 100k for me. GLM-5 was already getting pretty shaky at this point, so I feel like I at least have some kind of parity. But at least glm-5 was speaking & thinking with real sentances, I could keep conversing with it somewhat, where-as glm-5.1 seems to go from perfectly level headed working fine to all of a sudden just total breakdown, hard switch, at such a predictable context window size.

It seems so so probable to me that this isn't the model that's making this happen: it's the hosting. There's some KV cache issue, or they are trying to expand the context window in some way, or to switch from one serving pool of small context to a big context serving pool, or something infrastructure wise that falls flat and collapses. Seeing the window so clearly change from 200k to 60k to 100k is both hope, but also, misery.

I've been leaving some breadcrumbs on Bluesky as I go. It's been brutal to see. Especially having tasted a working glm-5.1. I don't super want to pay API rates to someone else, but I fully expect this situation to not reproduce on other hosting, and may well spend the money to try and see. https://bsky.app/profile/jauntywk.bsky.social/post/3mhxep7ek...

All such a shame because aside from totally going mad & speaking unpuncutaed gibberish, glm-5.1 is clearly very very good and I trust it enormously.

I wonder if running the compaction in a degraded state produces a subpar summary to continue with.

Have you compared it with using Claude Code as the harness? It performs much better than opencode for me.

General intelligence (not coding) comparison: https://aibenchy.com/compare/z-ai-glm-5-medium/z-ai-glm-5-1-...

The blog post has a benchmark comparison table with these two in it

Compared to Kimi 2.5 or Qwen 3.6 Plus I don't know, but I ran GLM 5 (not 5.1) side by side with Qwen 3.5 Plus and it was visibly better.

Hmm, three spam comments posted within 9 minutes of each other. The accounts were created 15 minutes ago, 51 days ago, and 3 months ago.

Interesting.

Hopefully these aren't bots created by Z.AI because GLM doesn't need fake engagement.

Feeling very much the same. Attempting to use it through Claude Code as a model it just completely lost all context on what it was doing after a few months and kept short circuiting even with the most helpful prompts I could give, outside of just writing out the answer myself. I really do not get the praise for this model.

Being "better than Opus 4.6" is not really something a benchmark will tell you. It's much more a consensus of users liking the flavor of an answer, rather than fueling x% correct on a benchmark.