I do not know why people even use Max or Ultra levels of thinking effort. Most of the time, i found its better to just run any of the frontier level models, with high or medium. Their capability beyond those levels is often diminishing returns, in exchange for double or quadrupling the costs (or usage).

And the bonus is that often at those levels, they tend to spin up less sub-agents that just eat away at tokens/usage, like its water in the desert.

A Max or Ultra is something that really only belongs if medium or high can not solve a very nasty bug or issue. Even in planning its often too much.

This keeps happening with every good new model release. People jumping from one to another, and as prices get lower, they start using the models even more.

People make not like to hear it but prices and usage limits are ways to shape traffic. The third option is the nuclear one like Kimi did, by just stopping to sell subscription at all. But that is something that DeepSeek can not do as all they offer is API.

Even OpenAI despite having the most compute is not immune to client influx = capacity issues. As people found their usage dropping, despite the push to the easier to run Luna models.

Reality is, that compute can not keep up with demand, especially when models get more capable and cheaper. What trigger people being using them more, what trigger compute crisis's.

This constant up and down cycle is going to keep happening for a long time, as this new market grows and eventually, somewhere in the future stabilizes. But yea, that is still going to be a few more years for sure.

Its in my opinion not wide spread at all and as today is literally the first time i ever hear anybody mention this.

Meanwhile the Chinese models are "go ham dude"...

If it was not for capacity issues, Chinese models have a higher change to just dominate.

> $2t company by the way

It used to be that OpenAI and Anthropic had such a moat around them, that such a valuation was worth it. But these days, its gross overvalued (like so many).

The more stuff is being pulled like cyber verifications, downgrading effort levels, downgrading usage (OpenAI), the more people move to those Open Weight Chinese models.

A fun recent event ... https://opencode.ai/data/

When DeepSeek Flash 0731 came out and provided a massive jump in cheap inference capability. It resulted in a 10x increased OpenCode token usage.

It took a 2.5x to 5.0x price increase AND a reduction by 4x usage (later to 2x) usage, and several cheaper models + a free model, to push the traffic down.

Traffic towards open weight models is increasing, even if providers can not keep up with the influx of new customers. This is not something you want to see as two companies, trying to go for IPOs.

So the idea of stonewalling cyber capabilities, when the rest of the world is just doing whatever with open weight models, on their own hardware even! This entire strategy from Anthropic never made any sense.

Most people do not realize how much of a shift that is.

In the past there was always a ton of EU countries helping the US out, in whatever crap the US started. And a lot of European soldiers died. Those soldiers did not help Europe but the US, because of being allies.

Even when some European countries got called names for not helping in Iraq because they called out that the whole Weapons Of Mass destruction was a load of bull** . It did not stop others from helping the US.

Yet, now, what we see as "help" is so minimal, and even countries like Spain not just refusing but active blocking. It shows how much relationships have deteriorated.

Trump thinks of Europa like a vassal state, and every time we say no or delay, it drives him nuts. Or our best "yes", if often a "well, we need to have agreement in all the countries, see you in a few years". Do not think its coincidence, its done deliberately.

In the past when the US said something, beyond maybe France, most countries quickly followed. This is gone ... We know the world has changed, and the old order is finished.

https://x.com/deepseek_ai/status/2090730032574631962

Vision + increase in benchmark results.

O, it can be even worse, when you realize its a 30 year old problem ...

During the 1998 television show Schalkse Ruiters, presenters Bart De Pauw and Tom Lenaerts dressed in fake pilot uniforms. They bypassed security at Brussels Airport (Zaventem), entered a Boeing cockpit, and left undetected to expose safety flaws

The TV show got cancelled not long after this incident because of pollical backlash.

The show had a reputation of finding security flaws (like being able to transfer money from people bank accounts) and other issues, but the airport one was the end of the show.

People loved the show, as it forced companies, ... make changes to their processes that they normally never did. They even did follow-up episodes to see if the companies actually made changes.

Bit of public shaming to fix security issues... worked great. Until the show was cancelled. The number one rated show of the country ... Yea, there was absolute no correlation between its cancellation and the airport incident. Really ;)

I think that Flash is still a usable model but Pro is DOA... Even before the price difference between Flash and Pro, vs the intelligence / problem solving / tool calling did not make sense. But now that gap has widen even more. And there are just too many competitors models now close to that Pro price range.

Especially when we compare that competitive models offer subscription services that easily cut down the token price by 1:10. That makes Pro especially a bad value.

We shall see what the 3th party market is going to do, but i suspect that prices will be increased. If the argument was that DeepSeek increases price as they lack capacity, a company with access to billions, other 3th party providers that need to rent and have less optimized infrastructures will increase prices. Especially if they get hit hard with people moving around.

Its like we always see the same issue with popular models.

* GLM 5.2 is good, capacity issues, API price up, subscription heavy nerfs. * Kimi K3 is good, capacity issues, API price up, subscription heavy nerfs. * DeepSeek V4 GA is good, capacity issues, API price up * OpenAI GLM 5m, 10m active users. Subscription usage is sneakily tightened more and more. * Anthropic Opus too popular, ...

That is the main issue. The AI users are people who actively easily move between companies. Pushing peak loads to each unprepared company, releasing load on the "less desired". And round we go ...

It allows for much more context that flow with your thoughts. Where as when you type, you tend to shorten you thinking process trying to get the bulleting points in, but that often ignores smaller things. And then you think "i can add this later", but that never happens because rabbit chasing the LLM.

So far all the suggestion that Opus 5.0 offered me, always aligned with what i wanted. Its not just Opus that i noticed this with.

User Posts can be downvoted but you need over 500 karma to have access to the downvote button. A Submission can not be downvoted.

In order to get high bandwidth, you want memory as close as possible to the GPU. The more trace lane length = signal loss, bandwidth loss.. HBM is compact, and so you can stack 24GB modules, 8 around a GPU die.

If you tried to do that with normal memory, you need like 64 modules. So a a TON of traces more that all need to be equal length, and because so many = far away from the GPU = less bandwidth.

The issue is like stated above, its a process that waste a ton of wafers. Wafers that can make easily 3x more normal memory.

Intel with "Crescent Island" is trying to make a 160GB card using LPDDR5x memory but the bandwidth is only ~700GB/s.

We are seeing stuff like AMD buying Taalas, with their Llama 3 8B on a chip, being able to push 17.000 tokens / second.

https://chatjimmy.ai/

Generated in 0,033s • 14.212 tok/s ...

People need to understand, that even AI companies do not like to build datacenters and the high energy needs. It cost them money and they are also looking at ways to develop better hardware / solutions that gives them more inference for cheaper (what means less power draw / heat generated).

What cost the most in API. Input, Cached Input, or Output. There you have your answer.

Unfortunately, we have moved so much of the actual intelligence of models towards reasoning, what results in some models getting good scores, but this is because they are dumping a insane amount of reasoning tokens at the problem.

So a mid priced model, with heavy reasoning output, cost the same as a expensive model, with medium reasoning output.

Before the GPT Luna price drop of 80%, you actually had the same price if you used Luna High and Sol Low. With the difference that Sol Low was insane fast, and often way better code.

https://deepswe.datacurve.ai/

Do not look at the top score but more what is on the horizontal axis as you go down. Sol Medium is frankly, was the best performance for dollar, until that Luna price drop. I will even argue that despite the higher price, Sol Medium is still way better despite Luna Max being cheaper. Or Opus Low, one of the better values also.

What do you notice? Is that those models all have a high intelligence start point for their low setting. So that means they do not rely as much on output tokens aka thinking.

Interesting quote by OP on reddit:

> So I wouldn't say "AI saved me $200k". What actually happened is that without it I'd have quit around month nine, like every other side project in my graveyard.

https://www.reddit.com/r/vibecoding/comments/1vgl5as/the_roi...

Comments URL: https://news.ycombinator.com/item?id=49190182

Points: 2

# Comments: 1