Ah, forgot to add, it's not really "unified" you have to explicitly specify your allocations. You may have a reasonably good 48gb chunk assigned to the GPU, but that DDR5 is 5-10 times slower than GDDR/HBM and the GPU itself isn't stellar.
So, framework laptops are great for chatting but nearly useless in agentic coding.
My Radeon W7900 answers a question ("what is this project") in 2 minutes, it takes my Framework 16 with 5070 addon around 11 minutes without the addon - around 23 (qwen 3.5 27b, claude code)
> It's largely SoftBank, Oracle, Microsoft and Nvidia, all of whom don't have big piggybanks full of hundreds of billions.
Actually SoftBank, Microsoft and Nvidia literally have free cash sitting there.
NVIDIA for example had over $60B in audited, reported free cash flow in 2025[1]
> loans from Nvidia (at high interest rates),
Is this just something you are making up?
"NVIDIA intends to invest up to $100 billion in OpenAI as the new NVIDIA systems are deployed. The first phase is targeted to come online in the second half of 2026 using the NVIDIA Vera Rubin platform."[2]
The closes there is to waht you are saying is reporting that NVIDIA has discussed guaranteeing some of the loans OpenAI is taking to build data centers:
"Nvidia is discussing guaranteeing some of the loans that OpenAI is planning to take out in order to build its own data centers, The Wall Street Journal reported, citing people familiar with the matter."[3]
This of course is the opposite of NVIDIA loaning OpenAI money - if they did this they would be liable for OpenAI's debts.
Just because they have cash doesn't mean they're going to lay it all on the table for one risky business. Which is why the actual solution is mostly loans.
In being a guarantor, Nvidia isn't directly giving the money now, but it is promising to be liable for the loan if OpenAI defaults. That's a risk they have to budget for.
But in addition to being a guarantor, they're also loaning out chips:
"OpenAI plans to pay for Nvidia’s graphics processing units (GPUs) through lease arrangements, rather than upfront purchases.""By leasing the processors, OpenAI can spread its costs out over the useful life of the GPUs, which could be up to five years a person said, leaving Nvidia to bear more of the risk."
Nvidia is leasing them the GPUs, which is basically a loan.
"In addition to offering a cost-efficient way for OpenAI to access chips, Nvidia’s lease option and long-term commitment can help the company land better terms from banks when it comes to raising debt, a person said. "
"As a non-investment-grade startup that lacks positive cash flow, financing remains costly. OpenAI executives have called equity the most expensive way to fund data centers, and said that the company is preparing to take on debt to cover the remainder of the expansion."
"“Folks like Oracle are putting their balance sheets to work to create these incredible data centers you see behind us,” Friar said. “In Nvidia’s case, they’re putting together some equity to get it jumpstarted, but importantly, they will get paid for all those chips as those chips get deployed.”"
Nvidia leases (loans) them GPUs with a promise to pay for them later. Getting this lease (loan) from Nvidia helps them secure more debt.
"OpenAI’s path to a $500 billion private market valuation has been enabled by hefty investments from Microsoft and others that allow the company to burn billions of dollars in cash while building its AI models that power services including ChatGPT."
"Jamie Zakalik, an analyst at Neuberger Berman, said the Nvidia deal is the latest example of OpenAI raising money that it pours right back into the company providing the capital. Investors are concerned about the “circular nature of this deal goosing up everyone’s earnings and everyone’s numbers,” said Zakalik. “But it’s not actually creating anything.”"
They get a loan of chips, use it to get a loan of money from someone else, and spend that money on the company that loaned them the chips. It's a Ponzi scheme.
This is just a grab bag of different critisms to what you said.
Vendor financing is very standard across multiple industries. It's not a loan in the sense you originally stated because it's physical assets, not financing.
The circular nature of the deals is a valid concern but different to what you raised.
It's also a much smaller concern now with Anthropic hitting $30B ARR from non circular sources than it was in September 2025 when your article was written.
> Plus I think most people using agents for coding are using subscriptions which they are definitely not profitable in.
Where on earth do people get this idea? Subscriptions that are based around obscure, vendor defined "credits" are the perfect business model for vendors. They can change the amount you can use whenever they want.
It's likely they occasionally make a loss on some users but in general they are highly profitable for AI companies:
> Anthropic last month projected it would generate a 40% gross profit margin from selling AI to businesses and application developers in 2025
and
> OpenAI projected a gross margin of around 46% in 2025, including inference costs of both paying and nonpaying ChatGPT users.
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.
Sure you did. But 99% of the time, you get the benefit of things that come with ability to sue - such as the vendor having a support team that's actually incentivized to respond to reports and deal with them quickly.
I agree with parent, "having a contract" gives you nothing tangible. Big tech providers get hacked quite commonly nowadays, some with glaringly embarrassing vulnerabilities like "the admin password was admin". All your data leaks, and the most you get from them going "sorry".
Having the ability to sue, and having the resources to sue is also not the same.
The amount of times I had to deal with support cases (as the reporter, not the handler) where I felt like the support person was actually incentivized to solve my problem vs just following the script is astonishingly low. Even with paid support. Paid support just means you get to follow their script faster.
So you don’t use any other open source software at all then?
The risk with OpenClaw et al isn't that the software itself is compromised. The risk is that what it does is fundamentally insecure and Claude Code isn't any better
That’s not the issue, the issue is that people are using their subscriptions (intended only for use with Anthropic products) with non-Anthropic products and this is simply Anthropic enforcing their ToS.
That's table stakes. LLMs are not like traditional software for fundamental reasons, and cannot be fully secured without destroying all value they provide.
Once again, despite everyone's protestations about not anthropomorphising things, LLMs are, to first approximation, best seen as little people on a chip. So with that in mind, it should be obvious why enterprise would prefer dealing with Anthropic's official products than OpenClaw - it's similar to contracting a team of software engineers from another well-known corporation and giving them keys to the castle, vs. inviting in any randos that show up at the door on any given day and can pass FizzBuzz test. Even if, in both cases, these turned out to be the same people, having an organizational/legal-level relationship changes the expectations and trust levels involved.
Both teams ship at breakneck speed and both randomly regress. I don't see such a big difference. Claude now uses Claude by default to judge whether a tool call is sane or not. At least OC is transparent about the insanity of running bash commands unchecked.
>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.
It's pretty clear that they do continually adjust the amount of tokens in a subscription, per se (and at best they offer sort-of estimates of quotas). The same activity exhausts my session quota on one day, yet it's a minor contributor on another. They make this very explicit with the "2x" event for the past two weeks, but anyone who uses it knows this is basically an ongoing reality: If you stick to using it off hours, you generally enjoy a more liberal usage grant.
But if they just "adjust the amount of tokens in a subscription", they would be punishing everyone for the outliers. The average normal user has spurts of usage where occasionally they need more and then there are gaps where they use little.
Subscription services rely upon this behaviour, and the economics only work if they "oversell". That's why OpenClaw users want to sneak in under a subscription, because the tokens come at a discounted rate over using the API based upon that assumption, but they are breaking the model because those users aren't conforming to expectations. It's basically the tragedy of the commons and a small number of users want to piss in the well.
> 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.
I think that's part of it, the other part is that OpenClaw is OpenAI IP now, and Anthropic want to allow users to ensloppify the internet through their own features now instead.
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).
I love that you are doing this test. However, as it purports to be a test of "English-to-SQL", your hardest question (Q9) seems ungrammatical:
> Show order lines, revenue, units sold, revenue per unit (total revenue ÷ total units sold), average list price per product in the subcategory,
gross profit, and margin percentage for each product subcategory.
In particular, the clause "in the subcategory, gross profit, and margin percentage for each product subcategory" is ambiguous, and I wonder if more models would pass if the English were reformulated to be correct.
(it's also notable that Claude Opus 4.6 and Sonnet 4.6 both "missed" this one)
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!
reply