I suspect people are misdiagnosing the root cause of why Anthropic is doing this a bit.
I don't think this is particularly about the financial impact of people using OpenClaw - they can adjust the amount of tokens in a subscription quite easily.
I think the root cause is that Anthropic is capacity constrained so is having to make choices about the customers they want to serve and have chosen people who use Claude Code above other segments.
We know Anthropic weren't as aggressive as OpenAI through 2025 in signing huge capacity deals with the hyperscalers and instead signed smaller deals with more neo-clouds, and we know some of the neo-clouds have had trouble delivering capacity as quickly as they promised.
We also know Claude Code usage is growing very fast - almost certainly faster since December 2025 than Anthropic predicted 12 months ago when they were doing 12-month capacity planning.
We know Anthropic has suffered from brown-outs in Claude availability.
Put this all together and a reasonable hypothesis is that Anthropic is choosing which customers to service rather than raising prices.
I'm at large company and pretty much everyone has settled on opus or sonnet 4.6. We would absolutely not allow something like OpenClaw on our network so your point kinda fits here where, if capacity is constrained, then by setting focus away from OpenClaw you're essentially prioritising the enterprise clients.
Just spitballing of course
They clearly see having a wide set of paying customers as valuable (otherwise they'd just raise prices) but if you are stuck having to make hard choice then I can see the attraction of this approach.
> not allow something like OpenClaw on our network
And where’s the difference between the Claude Desktop app and OpenClaw at this point? Anthropic have been hard at work porting the most important features. You can easily shoot yourself in the foot with both now.
I doubt anyone actually thinks the Iranian regime is good in anyway. But I thought the whole points of MAGA was "No new wars".
And now there's a new war, without any real reason (other than something something Netanyahu and they don't like the US) against a country that is a much more sophisticated adversary than Afghanistan or Iraq.
"sadly has been perfectly inevitable for decades"
Surely by now we know nothing is inevitable? Especially over the range of decades.
The naming is a bit odd - E4B is "4.5B effective, 8B with embeddings", so despite the name it is probably best compared with the 8B/9B class models and is competitive with them.
Qwen3.5-9B also scores 15/25 in thinking mode for example. The best 9B model I've found is Qwen3.5-9B-Claude-4.6-Opus-Reasoning-Distilled-v2 which gets to 17/25
gemma-4-E2B (4bit quant) scored 12/25, but is really a 5B model. That's the same as NVIDIA-Nemotron-3-Nano-4B which is the best 4B model I've found (yes, better than Qwen 4B).
Oh this page is great! I just released AIM [1] which is a tool that generates verified SQL migrations using LLMs, and I tested a bunch of models manually. I think I'll just link to your page too!
This non-Preview release scored 16/25. Probably the same model as the preview, or at least not particularly improved if you want agentic performance.
Good to see more options for large open models though!
It's hard to point definitively to a reason it underperforms
but generally models that perform well at agentic tasks were trained on very large numbers of tokens (Qwen, frontier models) or were heavily post trained for reasoning (see eg Nemotron-Cascade-2-30B-A3B at 21/25 vs the base model Nemotron-3-Nano-30B-A3B-Base at 12/25 )
Bit of a tangent, but I'm pleased to see that Qwen 3.5 35B is tied with GPT-5.4 and just 2 points behind 4.6 Opus. That little model is so impressively capable and fast! I'm frequently still surprised that I have that level of capability and speed running locally on my laptop.
Nemotron-Cascade-2-30B-A3B is worth checking out too at the size - I found it even better than Qwen 3.5 35B! I ran it slowly but successfully on a 8GB 1070GTX with CPU offload.
In the Dwarkesh podcast with Semi-Analysis's Dylan Patel they forecast the phone market will shrink by 50% this year because of RAM prices:
But that’s the high end of the market, which is only a few hundred million phones a year. Apple sells two or three hundred million phones annually. The bulk of the market is mid-range and low-end. It used to be that 1.4 billion smartphones were sold a year. Now we’re at about 1.1 billion. Our projections are that we might drop to 800 million this year, and down to 500 or 600 million next year.
We look at data points out of China from some of our analysts in Asia, Singapore, Hong Kong, and Taiwan. They’ve been tracking this, and they see Xiaomi and Oppo cutting low-end and mid-range smartphone volumes by half.
Yes, it’s only a $150 BOM increase on a $1,000 iPhone where Apple has some larger margin. But for smaller phones, the percentage of the BOM that goes to memory and storage is much larger. And the margins are lower, so there’s less capacity to even eat the margins. And they have also generally tended not to do long-term agreements on memory.
Why this is a big deal is that if smartphone volumes halve, that drop will happen in the low and mid-range, not the high end.
This is an extinction event for the low-cost cell phone companies. How are they going to survive if they can't sell their $100 phones profitably for 2 years? I think many of the low-end companies will simply sell their allocations of RAM and close up shop.
This is my greatest concern. So many small players will be wiped out. Consolidation is assured. Always great for consumers to be under the thumb of increasingly large companies.
There are a bunch of subbrands but there are also a lot of genuine small Android phone companies, especially in China.
Some of these serve some interesting niches that might now disappear due to this DRAM supply issue, e.g. Unihertz for extra small phones or CAT for extra durable worksite phones.
Will it be such a big deal though? Currently people are swapping out their phones for another model that's exactly the same but with a different number at the end of the name every 12 months. This could just mean that the unnecessary churn dies down a bit, and companies taking advantage of it have to find a new line of business.
> Will it be such a big deal though? Currently people are swapping out their phones for another model that's exactly the same but with a different number at the end of the name every 12 months.
I don't think they do that at the low-end (nor the high-end, though that doesn't matter here - higher-end manufacturers have a small margin they can eat into). People on the low-end phones want a new phone, they just cannot afford it!
Even in the mid-end: If you buy a phone which you find to be decent, but affordable, and are not out for chasing the latest gimmick - there is no reason it would not last you 6 or 8 years easily - before applications start assuming the presence of better hardware, or a newer Android version than you have etc. Naturally you will have to protect it from physical damage, and maybe replace a battery at some point.
Because the phones stop working well? I write part of a post, open another tab to go look up some information, come back to the post and what I've written is gone, because the memory got dumped. That's the reality of using an old cheap phone.
And would you consider yourself representative of the phone-buying public in general?
My desktop PC is from 2008 but I'd never consider this to represent anything like common usage. In fact it's so unusual that I get to point it out in posts like this.
This comment makes no sense to me. I exclusively use very low-end phones from Xiaomi. I buy a new one roughly every two years. Each new phone has a better screen, camera, CPU/GPU, charging, and sometimes more RAM/storage.
But it had 4k 60fps video, optical image stabilisation, a "super retina display" etc five generations ago. The specs have kept improving, but it's not a quantum leap in performance.
The same applies at the low end, the grand parent comment even agrees.
You buy a new phone every two years, it comes with a camera, a cpu, a gpu, a host of sensors. Same as phones did two years ago, and ten years before that.
I don’t use my current smart phone in any ways that are different to the iMate PDA2K I had twenty years so.
How often does your browser freeze up when you open a webpage? How often does your phone browser dump its memory when you switch to another tab and then switch back? Eg if you were writing a post and opened another tab to go check some fact then the post in the original tab gets deleted.
Because that's what happens if you use an old cheap phone in the modern day.
I even had a phone that would occasionally just crash when on a heavy website and the onscreen keyboard popped up. That was not at all infuriating!!! Especially when it would crash when I try to refine a Google search.
your comment makes also no sense to me, I exclusively use very low-end phones from Xiaomi since 6 years, and change it only when it's dead or when I can't run my apps (I'm afraid mine won't last 2 years more). Before this I kept my first smartphone (iphone 3GS) for 10.
Forget developing countries, iPhone is a luxury even in some European countries, when rent is 500+ Euros and your take home pay is ~1000. After all the other bills you're not left with iPhone money, which is why 100-200 Euro models of Chinese brands are doing so well.
It's easier to name the countries where iPhone ISN'T a luxury, as you can count them on very few hands.
Many countries would develop much faster if there weren't bombed nor maintain by puppet dictactors from (economically) developped nations (USA and france keep doing this intensively, while countries like Germany dont mind supporting fascist states). (PS: I'm not woke, not even Marxist).
Thanks a lot, Sam Altman / OpenAI. Their little $100bn war chest being used for obstructive / destructive purposes will wipe out multiples of that amount via economic ripple effects. All in an attempt to keep a stranglehold over AI via competitive resource starvation. Basic.
> This is an extinction event for the low-cost cell phone companies. How are they going to survive if they can't sell their $100 phones profitably for 2 years?
This is a great thing to happen, actually. Those phones are all essentially trash that ends up in a landfill within a year or so. They should not exist at all.
Smartphones are widely available on the used goods market though, perhaps even more so than second-hand SBCs or old PCs. The "low and mid range" can be filled by the former high end.
My Samsung Galaxy S3 died after 8 years. EMMC failure. Just started boot looping while I was asleep. Everything gone. Known issue.
My Samsung Galaxy S8 died at 7 years. Some kind of thermal failure, I was able to recover my data by keeping the phone in the freezer while I copied. Known issue.
My Samsung Galaxy S21? I figure I've got another year or two in it before it, too, dies.
Having beautiful dead phones that have never had a broken screen or a hard drop is pretty depressing.
It can also depend on the hardware it's connected to. If the endless gigabytes of Samsung's value-add software are scribbling to eMMC nonstop then it's not surprising the flash is wearing out. A lot of this stuff is masked by the fact that most people swap out their phone for a new one that's exactly the same every 12 months so they never notice this, but if you hold onto a phone or similar device for longer the unnecessary wear starts to add up.
Google Android should get more praise for doing quality control by analyzing and killing apps and processes that attacked the hardware - at least back in the day.
The great filter for incompetence by the big G was real and necessary.
Yeah, the Flash has a wear lifetime. Battery has a finite lifespan too. Anything over five years is pretty good going. My wife managed that with a Nokia 1020, the last and best of the Windows phones.
Like everything else, phones need to be backed up.
I just replaced my OnePlus 5 a couple of months ago at over 8.5 years old. No repairs needed, battery was a bit crippled in active use, especially making calls, but fine for a mostly idling phone. In idle it still lasts longer than a 1.5 year old iPhone 15. I still use it for by backup phone number SIM, as it slowly gets to ~9 years old.
The bigger issue was no more OS updates since 2020, and no Play updates since 2023. The battery can be replaced but getting a fully updated OS is more involved.
OnePlus 5 runs great with custom ROMs, including potentially ones based on mainline Linux as opposed to AOSP. (The Linux support is not as good as OP 6/6T but getting there pretty nicely.)
Too bad they have these long lists of "this doesn't work so well" and I'm too time constrained to troubleshoot for too long or dig for solutions. And I'd also need to replace the battery. It's an option for when I actually have some time.
The device integrity is a bigger deal, this is also a backup for some banking apps so if they don't work it kind of defeats the purpose. I removed all other apps to minimize the attacks surface.
If you're using it as backup for banking apps and the like I totally get not running a custom ROM on it! But you could also set that backup on something even cheaper, any one of the random not-bootloader-unlockable brands, and be left with the OP5 as a Linux phone. You're also right that the Linux support on OP5 is not up to standard yet, this is more of a question for the future if that support improves.
What's cheaper than an already existing phone that would otherwise stay unused or end up in the landfill (recycling center)? It could also be a great experimentation platform, play with Linux on the phone, but the time I have available now leaves little room for this kind of play.
The goal is not "experimentation" but having it eventually as an always up-to-date daily driver once the support for it matures. You're quite right that we're still a bit far from that, though.
>Smartphones are widely available on the used goods market though, perhaps even more so than second-hand SBCs or old PCs. The "low and mid range" can be filled by the former high end.
When new cars got more expensive, used cars got more expensive, too. I expect the same to happen with the phones.
If you're not a phone power user, you can get by on old low end stuff. When my pixel 4a died of a bad screen crack a couple years ago, I replaced it with a random used 4a on ebay for $80. Two years later and it's still completely fine for all my purposes (texting, phone calls, chrome browsing, tolerable camera, etc.), although I still haven't accepted google's deal for a free battery swap yet from sheer laziness. I've learned that I can accept a 90 minute screen-on phone battery, though it's an odd adjustment to make. Again, not a power user.
The free battery deal ended in January, but you're likely better off as mine ended up getting a damaged screen while being transported for mail in (because all local stores stopped doing the program), and they wanted to charge me an extortionate price to fix the screen. Support were useless
When it becomes clear that the insanely expensive AI data center orders are not going to be filled, we can expect a huge reversal in the price of RAM and GPUs. There are 241 GW of orders on the pipeline but only a third of that is under construction and of that third, even less is being quickly finished and brought online, It’s estimated that just 3GW came online last year.
Don't count on it. There's a lot of money in killing other businesses, or even just keeping prices high. Even if the high prices are an accident, there is always someone looking to take advantage of any situation for profit.
I have to agree. You only have to look at car and junk food inflation from after covid.
The prices make no sense, but that doesn't matter, they got away with it and are fighting to hold onto high prices, even as consumers balk. Their solution? Ditch poorer consumers. New cars and (branded) junk foods are luxury items now, apparently.
I had an iPhone 11. It was a good phone. It started giving up in early 2024. I held on with poor battery life until the new iPhones that year and bought the 16 Plus. I'm glad actually because they're discontinued the Plus models, annoyingly.
But I'm glad I don't need to upgrade for the next couple of years. I honestly want to get 4-5 years out of any phone going forward. There's basically no difference between models 12 months apart.
I can see the prices going up this year. IT's already happened to the PS5, which is bascially unheard of.
It really sucks more because the reason for it--AI--is just so godawful and pointless.
I bought my wife an iPhone 11 Pro Max in 2020, and, knock on wood, outside of replacing the battery it has been going on like a champ.
I've offered to buy her a replacement phone but at this point I think she's kind of curious as to how much life she can get out of it.
I have an iPhone 13 Pro Max; I bought it in 2023 but it was a refurb so I don't actually know how old it actually is. Regardless, it's still going strong, and I am hoping it can last through whatever RAM crunch is going on.
>But I'm glad I don't need to upgrade for the next couple of years.
Said the user who didn't learn the lesson.
Apple, you do not own anything, if Apple wanna release an update next month that makes your current phone useless, there is nothing you can do to prevent it.
Apple was caught hacking battery level, hacking users GPS signal, etc.
> AI is the best thing I've seen in 30 years working in software and expensive RAM for 2 years is a price I think is worth it.
I think generative AI is pretty neat, but I'm not sure it's worth the RAM increases. I use Claude like everyone else does, and it's cool, but I am a little concerned at how much absolute low-effort crap is being produced with it.
It has made YouTube considerably worse; there was already a lot of low-effort shit flooding it, but now it's almost cartoonish. A lot of the videos that I'm recommended will have thousands of views, and give kind of a facsimile of a video with "effort", only for me to realize about a quarter of the way through a bunch of AI tropes in the writing and/or the visuals. It has made the already-mediocre experience of YouTube actively bad.
I am also not convinced that the prices will go down after two years. We already have big memory vendors completely leaving the consumer market, and we have these AI companies buying literal years worth of entire production lines of RAM chips.
This is something that could be solved by competitors jumping in to fill the niche, but it takes a lot of time to build new factories for this stuff, I think more than two years.
And if the AI bubble is still going then with datacenter construction going as planned, the RAM shortage will be even more extreme than today despite higher production capacities.
The only two changes that matter to me are: no more iPhone Mini and no more hardware switch to mute the phone. Instead they got a new "thinnest iPhone ever" that's actually thicker than my 4+ year old one when measured honestly.
I would not call any Apple phone even mid-range, and certainly not low-end?
This coming from an android user, that recently bought iphones for my daughters. Paying 600$ for not even top-of-the-line phones does not scream low-end to me.
My company-provided work phone is a base model iPhone, I'd definitely put it in a performance class lower than a flagship from any brand. Certainly not low-end, but I think mid-range would be a fair characterization.
Or optimize the os because I still find 8GB insane for everyday tasks. Ok, gaming I can understand, but most common tasks should be runnable with at most 2GB of memory and that is mostly for browsers.
Its going to be interesting, when big the big AI bubble directly attacks via pricing the government by preventing the sale of surveillance devices. So the bubble will pop- not because it can not sustain, but because its existence is adversarial to government demands for surveillance.
I don't think this is particularly about the financial impact of people using OpenClaw - they can adjust the amount of tokens in a subscription quite easily.
I think the root cause is that Anthropic is capacity constrained so is having to make choices about the customers they want to serve and have chosen people who use Claude Code above other segments.
We know Anthropic weren't as aggressive as OpenAI through 2025 in signing huge capacity deals with the hyperscalers and instead signed smaller deals with more neo-clouds, and we know some of the neo-clouds have had trouble delivering capacity as quickly as they promised.
We also know Claude Code usage is growing very fast - almost certainly faster since December 2025 than Anthropic predicted 12 months ago when they were doing 12-month capacity planning.
We know Anthropic has suffered from brown-outs in Claude availability.
Put this all together and a reasonable hypothesis is that Anthropic is choosing which customers to service rather than raising prices.
reply