Hacker News

You made me realize something. I routinely spend upwards of 500$ per month on LLMs for coding (expensed towards clients). However I live in a place where 500$ is around the avg. salary. I’m lucky that I know my way around western clients. Clients who pay these expenses and are happy to work with me because I am still about 50% cheaper than local talent in EU/US, while my salary at home converts to an upper class income at the highest tax bracket.

Which of course causes some unfairness on both ends. Nobody here can compete with me. I often use left over tokens on local client projects; which despite lower pay, still pays off because they now take hours not days or weeks to complete. And nobody in the local clients talent pool can compete with me; unless they charge about half the market rate.

Take away my 500$ monthly grant; and I’d be more or less screwed. Better open models will more or less start to reduce this advantage. It’s not like I positioned myself here on purpose. But it’s definitely a „right place, right time“ situation.

If we can agree that the AI model is at least as capable as a junior engineer or new contractor, how’s that different to saying “software engineering isn’t worth $200 a month”?

Has a very race-to-the-bottom feel to it.

Though in the grand scheme of it, $200/mo probably isn’t the real price either. Also looking at it not just in a vacuum - paying for a product that can change what you get from under you doesn’t seem great anyway.

At least with a locally-hosted model you know what you’re getting.

Yes, but you’re paying with your data unless you’re hosting with a provider you trust or self-hosting.

DeepSeek through their own API has saved me tons of tokens honestly. Even though it is not as smart as Kimi or Claude, their level of entry is very low with a top up of 2$ and Pay as you go compared to the subscription of Claude or 20$ top up of Kimi

As much as I don't like Mark Zuckerberg, part of me wishes he would get his head in the game and compete with these models, he's literally got all the capability to do so, and he could easily sell the model through deals with GCP, AWS, and Azure. Hell, Amazon needs a hot model they can host that's exclusive to them I feel like, maybe he can work something out with them, whatever the case, it seems so glaringly obvious to me, I'm not sure why he hasn't taken a stab at competing with Claude Code or at least frontier open models and then cutting a deal with cloud providers to recoup the costs of maintaining said models.

He's sitting on a frontier model letting it burn a hole in his wallet that could actually pay for itself.

Someone else on this forum put it well, U.S. is trying to achieve AGI at all costs, while Chinese models are seeking widespread adoption.

Significantly cheaper than comparable models if you are using openrouter [0]. Just yesterday I spent roughly 13 cents centering some divs using Deepseek in a personal project. It would have been north of $1 to do that with a US frontier model.

0. https://openrouter.ai/compare/z-ai/glm-5.2/anthropic/claude-...

The tokens cost the same everywhere on earth. This does hurt some cost advantages of outsourcing when tokens start to become a bigger part of development costs.

I read these stories and I can never figure out how people are managing to use these $200 plans. If I really go full bore, I can sometimes max out the $20 plan. Even then, it already produces more code than I can reasonably review and merge.

> It's increasingly feeling, to me, that theres a gap building up between haves and have nots.

People speak of a permanent underclass.

https://www.nytimes.com/2026/04/30/opinion/ai-labor-work-for...

With open weight models there is true inference competition. Whoever can serve the model at the lowest price. And the consumer wins. Capitalism, served by China.

Just don't ask it to tell you the events of June 4, 1989.

200 is much less than the value you’re supposed to get out of it. If it’s not then yeah go ahead and use cheaper models with worst quality

I also use DeepSeek v4 flash and v4 pro, but I can’t settle between using Claude Code or OpenCode and it seems like I waste time switching back and forth (especially keeping my personal SKILLs files synced). On one hand, a ton of engineering work has gone into Claude Code, on the other hand all Chinese models I have tried with OpenCode seem well configured out of the box.

I was thrilled to have Gemini Ultra for a month and use as many Opus tokens with AntiGravity as I could use, but I am happier using less capable models like DeepSeek knowing that it is more fun to do more of the work myself, it is a smaller hit on the environment, and incredibly cheaper.

I tried Deepseek V4 Flash with very low expectations and was pleasantly surprised. It's a surprisingly capable model for the price.

That v4 quality is available to everyone in the world for a pittance is beyond remarkable.

I use Pro because I’m insensitive to the price difference, but also found Flash very capable in OpenCode.

it would be a really great option if it didn't lack vision

Kimi and GLM models have coined a new term: Thinkslop. They run a chain of thought that is up to 10x longer than other models and it seems that through a lookback mechanism they are able to use the CoT to reason about solutions to tasks they couldn't otherwise solve.

The downside is of course that they consume many more tokens off your plan, and also that they are significantly slower. Kimi K2.7 takes about 7x longer to finish the same benchmark tasks as DeepSeek V4 Pro on my router benchmarks (https://role-model.dev/).

So for now I'm happy with just two models: GPT and DeepSeek.

I gave it my standard:

"Make a pac-man game in a single html page"

It went off and argued with itself for 20 minutes about how to lay out the map and then timed out.

> The model is good, the plan is a scam

If it is needing to generate that many tokens to do the same tasks, then it probably has higher inference costs. So (for you) the model is bad, the plan is the same plan.

What kind of tasks have you been using it for?

For me it works fine outside of afternoons Beijing time on weekdays.

Even as a GLM z.ai fan, I wouldn't pay for their plans. They are just way worse values than gpt or anthropic plans, in terms of both usage and capabilities.

Opencode Go subscription has served me well.

Self-promo but you should try our service synthetic.new. We generally have up-to-date open-source LLMs on the sub, and we have GLM-5.2 :) Perf+stability should be wayyy better than zai.

same here. Barely usable due to API connections issues.

And when i can use it, it just drains the quota 5 times faster than codex or claude.

Their plan is a scam

My experience as well unfortunately :(

> What provider do you use?

1. My own harness + Local (which usually means Qwen3.6-35B-A3B), I use this fairly often for research gathering on topics, info gathering on code bases, etc.

2. My own harness + DeepSeek v4 Flash served by DeepSeek, I added $20 quite some time ago and somehow still have $18.77 in there after I don't know how many prompts. I use this pretty often, slightly less than my local setup, it's great and what I'm planning on running locally (eventually).

3. My own harness + OpenRouter with whichever model I want to try out. I use this very rarely.

4. Pi + OpenAI Codex $20 subscription. I don't use this almost at all anymore, but I keep the Codex subscription for testing things out to see how GPT-5.5 will handle a problem the other setups have issues with.

> Why do you trust it with serving full quality?

The only thing I've noticed seems unbearably useless sometimes versus what I noticed before was GPT-5.5 which has had some of the weirdest degradations I've seen. It's not to Anthropic levels but it definitely had some service issues a few times where I was wondering if they had accidentally (or purposefully) lobotomized it.

Everything else has mostly just been the same, except DeepSeek I noticed had some speed issues a few days ago.

> What harness do you use? Why do you trust it not to have malware (most harnessed are TS apps)?

I pretty much only use my own, agents are trivial to make and it's definitely not hard to make one that's better than Claude Code or Codex for whatever you're doing.

Z.ai legacy Pro coding plan which will last me until the end of the year + maki.sh as the agent.

OpenCode works fine, i just find it very resource intensive for no good reason.

Next to my Claude Pro plan, I have subbed to OpenCode Go. I find the OpenCode UX much better than in Claude Code CLI. As for models, I started a few months ago with GLM 5.1 and it was solid and could archive near sonnet-level tasks. It weirdly sputtered out Chinese characters sometimes. Then I switched to Kimi K2.6, which is the Chinese model I used the most until now. It used way too many reasoning tokens (improved in k2.7). But executed Claude created plans reliably. Now I’m back with GLM 5.2 and it’s really solid (among other things it’s good at design) and I get good usage with the $10 plan. Still the Claude models have less hiccups but the Chinese models are getting really close.

    > What provider do you use.

    > Why do you trust it with serving full quality?

    > What harness do you use? Why do you trust it not to have malware (most harnessed are TS apps).

I use both the openai subscription and the opencode go subscription. I use the go subscription for my personal work and the openai subscription for my consulting work.

The differences between the models are minimal, but I usually stick with gpt-5.4-mini, gpt-5.4, mimo-pro-2.5, deepseek-v4-pro. These latter ones have way more usage than even using 5.4-mini so I tend to use them in personal projects for that reason.

My harness is https://github.com/can1357/oh-my-pi. I trust it...enough. It updates very frequently so as a safe guard I run it sandboxed with https://github.com/containers/bubblewrap so it can only access the project folder and some whitelisted config files

Pi is great, set it up with a system prompt to give the model more direction and think less, and it crushes anything I give it

For work, I mostly use Codex and some Claude. For personal use, I’ve started using Chinese models directly through their respective providers, mostly for automation tasks and experiments so far, either via the API directly or through the Pi harness.

I do not trust any of them. Everything runs inside virtual machines, not just the sandboxes provided by the harnesses. I also do not run Claude or Codex directly on the host machine. Not just because of supply chain fears, but also because of how incredibly user hostile the VC funded companies are when it comes to installing random stuff on your machine.

Synthetic.new and Claude Code using GLM-5.2. Great model, but the harness will error out if using subagents. The base plan only allows one concurrent request at a time. Also, GLM will burn through your weekly quota in a day if you're not precise with your scope.

Local using Qwen3.6-27B; 2xRTX 5070Ti graphics cards; VS Code with Cline at the moment and Ollama back-end (will get to trying the others soon).

GLM 5.2 coding plan- I'll post the agent as soon as I can! But opencode works and their own zcode is really good as well.

I don't understand those takes.

Open-weights perhaps, but definitely not self-hostable – since those require $20k+ capex – which is the real "step change" to me, as it ends the stranglehold providers have over censorship.

The only silver lining would be increased competition in API providers of those open-weight models leading to truly affordable prices and a race to remove stupid "safety" checks.

> What _is_ Google DeepMind doing

I feel like it has been pretty visible about what’s happening, between their press and products and financial statements. It’s just not what people are accustomed to expect.

First, Google has become a major compute provider for competitors, thanks to TPUs. They’ve talked about allocating TPUs to GCP instead of their first party products. I can only assume it’s because they’re collecting a higher margin, and it covers the cost of data center buildout - which they’ve been aggressively doing. I wouldn’t be surprised if they made the financial decisions to delay or slow training for Gemini 3.5 when they provided last minute compute to Anthropic this spring.

Second, Gemini has very directly not been focused on agentic coding, maybe 3.5 Flash being the change. They’ve built models they can deploy to watch YouTube videos, Nest cameras, scale to AI in search, understand fitness info in Fitbit, etc. They’re very clearly not focused around agentic/coding. They’ve put in a ton of efforts into multimodal data in and out, and they’re the only major lab working on video generation still. There was leak/rumor that their cofounder (brin) was getting involved in the model training to renew focus on agents so maybe this will change, and again 3.5 already feels different.

Just waiting for the 3.5 Pro they said would come out this month. Gemini is pretty much useless for any serious work right now.

copying the graphs and tables to HN is noisy and harder to read

> Will they still rent out their own model, will they support the open model and become a resource provider?

Anthropic rents GPUs from xAI to run Claude. If there's an open weights competitor to Opus, why wouldn't Elon host it directly?

Did you read the article? Opus 4.5 has essentially been rebuilt already

Anecdote, not “research”:

Yesterday I compared Deepseek, Kimi 2.6, MiMo 2.5 and GLM 5.2 for the same task (replace a custom token-based auth scheme with a cookies-based scheme across a front- and back-end codebase).

I used Opencode with the zen subscription to try different models.

All did this perfectly, basically indistinguishable from each other. However, when I pointed out that the new cookies-based auth didn’t allow multiple independent logins across browser tabs (which the previous scheme did allow) I noticed this:

Deepseek, Kimi, MiMo started giving me multiple options but advocating strongly that I should either accept this deficiency, or don’t use the cookies version (keep the old auth scheme). They were so similar it was as if they were all the same model.

Only GLM 5.2 said “here’s how to use cookies and also have tab-level separation”. The difference vs the other models was very stark.

Agreed. Even just distilled versions of their frontier models would be a huge win for the open ecosystem

The reasoning traces always look terrible and they’re frustrating to watch. It’s the same with Kimi. What’s interesting is that the end result is then good. I think it’s just some sort of devils advocate trick to get better output.

I have a hilarious theory why GLM (and Kimi) have this thinkslop,

apparently Chinese language as token is more information dense than English, so having these wasteful thinkslop in Mandarin isnt that damaging. So the developer focus mostly in Mandarin and didnt think of handling these thinkslop while American AI labs do.

I think the self-doubt might actually be a very crucial part of it's capability. I often feel compelled to interrupt when I'm watching it think (which thank the stars it let's us do, unlike the big American models!!), but usually it makes the right pick!

Being willing and able to reconsider seems very good. Going around and around, pulling in more thinking, integrating it: maybe that's why it is as good as it's good.

I want to emphasize again how excellent it is that we can see the thinking. I think this makes GLM so much better an experience for me. It gives me such insight into what is being considered, helps me see where things go wrong. It grounds me, gives me the notion of where the results come from. It was so jarring to switch to GPT and Opus and find that they won't discuss with me, won't reveal their thinking: that feels fundamentally unsafe, for me, for society, to have such a severe black box. I don't think it should be allowed, honestly.

Many thanks to this recent submission, which is the first time I've seen anyone blog about this core difference: The text in Claude Code’s “Extended Thinking” output is not authentic. https://patrickmccanna.net/the-text-in-claude-codes-extended... https://news.ycombinator.com/item?id=48630535

GLM-5.2 is a huge model. I don't think it would fit on the AI HAT+ 2 even if you quantized it to 2 bits

ime DeepSeek v4 Pro is great for cybersec/netsec, I have not tried GLM though

Now that's a tremendous pointer, I'm going to have to try that.

Do you full on let GLM5 get stuff done on its own or is it more like a guided workflow? The former's what the point releases doubled down on and is also something that uses a lot of juice.

>Right now it sounds like the US's export ban is not slowing them down a whole lot.

Just costing them a lot more money as they pay multiples more buying on the underground grey market.

> Right now it sounds like the US's export ban is not slowing them down a whole lot.

It may wind up being a massive boost to them in the long run, even.

Necessity is the mother of invention.

There does not seem to be a big penalty for going slow anyways. People seem to just switch on cost as soon as a model can do a task well enough. There do not seem to be strong network effects or vendor lock in.

Seems to me that going slow is the better long term tactic. China can just let the USA pay the high R&D costs to figure out what works, then just copy what works.

With subsidization from the Chinese government they will probably be equal to or better than the models here. I mean, have you looked at the author list of any given AI paper published within, say, the past 5 years? I wouldn't be surprised if half or more AI researches are from China.