Discussion summary
Discussions highlight the importance of tools for LLM usefulness, the availability of GLM-5.2 from multiple providers mostly outside China, and concerns about hardware and model migration in AI development.
What the discussion says
- Good tools are essential for LLM utility.
- GLM-5.2 is accessible from many providers, mainly outside China.
- Migration to open weights models increases flexibility.
- Hardware limitations may impact future frontier models.
- API strategies can hide proprietary technology.
“LLMs are quite useless without good tools.”
“Most GLM-5.2 providers are outside China.”
Join the discussion
Write your take first — we'll ask for email only when you're ready to publish.
- Hacker News
- I also think we’ll approach a point where increasing intelligence is not really going to suddenly improve most work tasks. I bet that’s already happened actually.
We’re oohing and aahing about models, when the ones a few versions ago did a good enough job for most of the dumb coding, etc we do
by softwaredoug - I don’t think this is true. All the models prior to Fable were honestly dumb as rocks, and Fable is too sometimes, but at least it’s helpful now and not a hindrance.
The future of AI most definitely involves making something twice as good as Fable that is virtually its own employee, and not on reducing inference costs because to be honest Fable isn't actually that expensive.
The real utility behind an AI model (imagining that it can be made twice as good as it is now) would be being able to scale a small business up and down instantly without hiring (to implement a new feature or whatever), which is costly and time consuming these days.
by aetherspawn - I’ve been on a GLM coding plan since they launched ~year ago and it’s been at „good enough“ since the start. Tangible behind absolute SOTA but like you say most coding isn’t rocket science.by Havoc
- The thing is they are inventing new things people will want to do. But for example, "loops", fully hands-off agentic coding etc., seem really unlikely to get much traction because that just isn't how software is designed within its producer/user community.
Requirements evolve in use, and fully hands-off LLMs simply cannot be trusted to only change the things you ask them to change, so I don't think it's likely that products will, in the main, be developed that way.
And if you don't need that fully-hands-off stuff, then the models that run on at least reasonably modest desktop hardware are surprisingly close to being enough.
by dofm - Seems like a pretty pointless post that still centers around output tokens.
In agentic coding, cached input tokens is 90% of the API "cost". It doesn't require GPU compute, and DeepSeek has shown that it can be done 50~100x cheaper with MLA/CSA/HCA, and a whole bunch of disks. This should collapse the margin.
by throwdbaaway - > MLA/CSA/HCA
Aren’t these techniques all “lossy” compression, and one of the reasons people complain about loss in quality as the context size grows larger?
- The current top comment in https://lobste.rs/s/ua1gxl/glm_5_2_coming_ai_margin_collapse correctly zoomed into cached input tokens, but landed on the opposite conclusion:
> That is, for your $100/month fee, you get $3600 equivalent of API usage. This is presumably because Anthropic has figured out some clever things to do with model routing and input caching, and also can subsidize with investor money and take a hit on their operating margins.
My take: this is exactly what Anthropic wants everyone to think. In reality, 90% of that $3600 are for cached input tokens, that can be made to cost next to nothing, as shown by DeepSeek.
by throwdbaaway - Aren't the American AI labs desperately struggling to find a market beyond just agentic coding?by nozzlegear
- Metaphor i like is that it will be as cheap as electricty?
Do you know who is supplying your electricity or which factory it runs on? probably no, bc its a commodity and mostly settled and there is so many energy resources. some are alternative some are coal mines. And they all fight in the supply demand trade for energy which is happening real time ( think open router here)
And eventually the consumer wins bc of the abundance.
I think greatest example of abundance of cheap infinite intelligence will be not glm5.2 but DeepSeek V4 Pro max with $0.435 per 1M input tokens and $0.87 per 1M output tokens
by Obertr - The metaphor breaks quickly because as a first approximation the electricity "quality" does not depend on the provider, and will not change overtime. That's not true of LLM output.by cdavid
- >the least understood upcoming shift in AI economics.
Then proceeds to talk about something in the AI news every day. Hey, did you guys hear? Open source models are cheaper and their quality is increasing!
So, first, by no measure is GLM5.2 as good as Opus.
Second, yes, open source models will put pressure on margins...eventually. Everyone knows that. But do you think today's AI business model is the same as tomorrow's?
- What if your measure is cost? Or zero-data-retention? Or diversity of inference providers?
On those measures it is better.
by james2doyle - re: "So, first, by no measure is GLM5.2 as good as Opus."
I accept that for you and your work this is true.
I have a different experience: for a month I paid big money for Opus and got a lot done. Now I am gorging on GLM 5.2 running on Fireworks.ai and I am also getting a lot done for about 15% of the money.
Everyone should do their own evals on their own work.
- > So, first, by no measure is GLM5.2 as good as Opus.
That's an opinion many will disagree with. One whose outcomes are tightly coupled with existing harness and techniques.
In my real life usage Opus 4.7 and 4.8 have been increasingly unhelpful compared to 4.6 in behaving as assistants.
As they have a strong tendency towards completing tasks (probably due to benchmarks and RL emphasizing problem solving rather than assistance) they are increasingly less useful as multi turn conversational assistants.
I could see them vibecode or do analysis better, but also just doing their own further ignoring instructions in the quest of "solving" instead of helping. Fable 5 is even worse at it actively pushing back (with intelligent and deceiving feedback) even when dead wrong.
GLM seems to suffer less of this.
by epolanski - > So, first, by no measure is GLM5.2 as good as Opus.
Depends what you do. Complex tasks, poorly-defined tasks, sure. For relatively simple tasks, though, or very well-defined tasks, it's just as good and usually a lot faster. It also has a more neutral character and is somewhat less adversarial than Opus. (Opus is always "Let me push back on that..." whereas GLM is "sir, yes sir!") I use both and I appreciate both. If Opus disappeared tomorrow, though, I wouldn't cry -- I'd be able to adapt to a GLM-5.2-only life real quick.
by A_D_E_P_T - GLM-5.2 is not as good as Opus, it's better. I can abliterate GLM-5.2 and have it work on projects that Opus refuses.by tarpitt
- > Of course, this was a hugely poor read of where the costs actually lie in AI. Training - while no doubt capex intensive - is a fixed, up-front cost. You spend hundreds of millions to train a model, then you are "done".
I don't understand this point that people make. If you're consistently needing[0] to train new models and the cost of training relative to the % improvement seems to go higher, isn't this just a constant cost that you continue to bear? The footnote seems to allude to this, but then sort of waves it away anyways. Also are there continuing incremental training costs to keep models relevant? Or do they only have knowledge of events up to the day they were trained?
[0] needing, because you have competitors and people expect more and more.
by dbalatero - Anthropic has essentially trained one big model for more than one year.
From Opus 4 to 4.8 all improvements were in RL and post training. Expensive, but not as intensive.
by epolanski - These models rely on knowledge that are embedded in their weights, if a new library is released, a new linux version comes out, some new protocol succeeds the previous one, you want your llm to know about it. Sure you can just add that into the context window, but that has its own problems.
Unless new research, there are a few which look promising, gives a new method, training is going to be a constant cost sink.
On top of this, if you stop training, it is 6 months until someone releases an open weights model and now you are competing to give the lowest price for the same product.
Also we can't forget that this is a business that *has to* be in the global labor industry, not just a tech tool, they have to have much better models to justify the trillion dollar evaluation
by blourvim - Last month, I cancelled my Claude Pro subscription and instead used those 20$ to purchase Openrouter Credits. Most of my knowledge-seeking questions can be answered by Gemma4, for basic code editing, Qwen3.6 27b is enough, and for really difficult tasks, GLM5.2 doesn't leave me hanging. I'm by no means a heavy AI user, so I'm even saving money going the API Credit route and relying on the smallest possible model depending on the task complexity.by pixlmint
- What do you use as interface to OpenRouter? I, too, am looking into using an API to see if I can reduce costs (I use OpenAI + Github Copilot, currently). TensorX instead of OpenRouter (because it's in Europe, and EURouter wanted 15% more money from me :P), but I'm not sure if I want to change a configuration in vscode every time I want to switch the model in the Claude extension (and having an API key in my settings feels iffy too >_>)by NostraDavid
- I literally burned through 20USD in a couple of hours on openrouter with deepseek v4 pro and opencode tasks - i'm sure i did something wrongby _def
- Unlike the belief that frontier AI is expensive due to a high margin, and going to be expensive if there is no competition. My understanding is that, under certain circumstances (which is most likely true), the price will be driven down just because of profit seeking.
The frontier LLM labs run on a huge fixed cost and very low marginal cost. They need the economies of scale to make sense of the business (an incentive to expand their user base as large as possible). Imagine that you want to buy a few B300s to run GLM 5.2 and rent the service out to other people. How could this business be viable and sustainable in the first place? You need as many customers as possible. If you charge everyone $1000, you find fewer customers who can afford it. It rots the ROA if the servers are not utilized 100% (you would better buy less compute instead).
Also, the marginal cost for onboarding a new customer is low. And it's getting even lower when you have more customers. You wouldn't leave money on the table (especially for your competitors) if you want to maximize your profit.
By this logic, all frontier AI labs are incentivized to lower the price to maximize their customer base, profit, and ROA.
by typ - I agree, but there is prestige to consider. Many people are motivated to buy the best, even if it's much more expensive. "We're building a mission critical application here. Sure the API costs are much higher, but it's worth it."by Gareth321
- > The frontier LLM labs run on a huge fixed cost and very low marginal cost.
> Imagine that you want to buy a few B300s to run GLM 5.2 and rent the service out to other people. How could this business be viable and sustainable in the first place?
My understanding is the frontier labs have huge fixed costs and relatively low marginal costs because they have to bear the cost of training the model/R&D, and then amortise that cost over their userbase.
By contrast, if I buy a few B300s and run GLM5.2 and rent the service out to other people, I can be profitable at a comparatively very small scale because I got the model for free.
by nr378 - It’s important that none of these entities can collude to price fix. Having China be the competitor ensures that.
Basic microeconomics is still the easiest way to understand token economies. How is it not a competitive market (where profits go to zero?).
Anything A or O does to keep more margin, any competitor can copy or choose to undercut, and undercutting has the benefit of collecting training data. So what is going to stop gross profit of tokens going to zero except for collusion/price fixing?
by spyckie2 - Considering conditions within a single market is still microeconomics, I agree though its tough to see where firms will get market power from so profit will tend toward zero. I thought the same about GPUs though and nvidia still doesnt seem to have any real datacenter competition in sight.by HDThoreaun
- You left out the one that will: federal government industrial policyby intrasight
- They have a vision MCP to make up for the model itself not having the capability natively: https://docs.z.ai/devpack/mcp/vision-mcp-server
I also found their web search to be mostly okay.
Furthermore, in case this is of interest to anyone, if you use their ZCode harness then you get bigger Coding Plan quotas: https://zcode.z.ai/en
Used it for a bit, it sits somewhere between OpenCode Desktop (still new but nice) and Claude Desktop (recent versions are good).
As for GLM 5.2 as a model - with max thinking it’s generally satisfactory, somewhere between Sonnet 5 and Opus 4.8, better than DeepSeek V4 Pro for sure.
Pricing wise, the subscription doesn’t seem as good as expected. I spent like 60% of the weekly limits of the Pro (50 USD) plan in one day, only because each 5 hour limit only gave me 20% to spend, otherwise it’d be 80-100%. Not even doing anything crazy, just parallel long form work on 2 projects with about 96% cache rate and at most 3 parallel code review sub-agents.
Their Max (100 USD) subscription would last me the whole week, but so does Anthropic for the same money and so would OpenAI. Off-peak is more palatable but I can’t just twiddle my thumbs at 9 AM to 1 PM local time.
Proper savings would show up with the Max plan and yearly billing, but that’s more of a tough sell.
by KronisLV - I switched to yearly Cline pass because it was too cheap hahaby zackify