I've had my eye on the normal Mac Studio M5 Ultra (normal plebeian 96GB RAM) and the release (edit: ship) date there keeps slipping, now into January. I'm surprised Apple is pushing the backlog even more.
I have one. It can run lots of smaller models really well. I’ve been loving Gemma4 especially. My needs more more large scale information extraction and not trying to replace Claude code locally though.
If anything the interesting models that can run in about 64gb of memory are increasing because most people can’t access more memory than that. I’m especially excited for clef and cleric as they continue to improve. Yeah, you don’t get 1mil token context length, but even frontier models can lose focus at that level
I've been holding off because I will need to do experimentation with local models at some point...but not yet and won't be for a bit but I am afraid prices will rise given component cost trajectories.
I've been tempted to just get the M5 Max instead and just innovate with smaller models myself.
> Sidenote: I should also add that you can see this is an anomaly when you use Anthropic’s own “dollar of revenue per compute” number. For Q2 2026, it brought in $2 of revenue per $1 of compute spend, up from $1.39 in Q1 2026. This number dropped to $1.70 in Q3. I think it’s a stupid number as it doesn’t really say very much about the business, but even based on Amodeiconomics, this quarter feels like the exception rather than the rule.
...making money is a pretty important part of the business?
Was just wondering if I can quit my job and live off of llm agent managing a fleet of vending machines for me. Curious is anyone here doing this? What's life like? How do you find meaning and purpose?
Makes sense this event is in LA- this this is designed to be a mobile media creation device. Most users don't care about most of the features that are highlighted in the press release, but someone creating any sort of visual media would.
Their demo example for the touch interface is piano keys? That's a 2010 Scott Forstall level of skeuomorphism that doesn't translate well for a laptop screen because it's not convenient.
The hands are also almost in the exact same image Steve Jobs had on a slide during the iPad keynote showing touch on a laptop is not ergonomic. It's poetic, almost.
Jobs routinely shit-talked product features and categories that Apple later went into, including both phones and tablets. People complained about it when he was alive, but now we’re asked to fondly recall how consistent he was?
"We've done tons of user testing on this, and it turns out it doesn't work," Jobs said at the company's "Back to the Mac" event in 2010. "Touch surfaces don't want to be vertical. It gives great demo but after a short period of time, you start to fatigue and after an extended period of time, your arm wants to fall off," Jobs continued. "It doesn't work, it's ergonomically terrible."
Well my arms didn't get longer, that's for sure. Also MacOS and software made for MacOS doesn't optimize for touch targets unlike something that will be distributed on iOS/iPadOS.
Not that novel. The first ergonomics studies on touchscreens were done in the seventies. In 2010, touchscreens were already sophisticated technology and well established in many settings.
I got my first laptop with a digital pen and touchscreen in 2008 or so. They were rare back then, but nothing groundbreaking new. The touchscreen did suck though.
I don't know why you got downvoted, it's true. A lot of screens were touch screen even in the 90s, but they were typically single-touch rather than multi-touch capacitive interfaces. They were used a lot (and still are) in Point of Sale terminals, for example, where an employee needs to stand and ring up sales.
This is the one thing I'm rather excited to use the touchscreen for. Would be genuinely useful to me right now. Beyond that though...... I can't see using Touch much.
I hope this sells well enough that they might later go the EXTRA step to a Pencil supported 360-hinged MBP.
I'm a hardcore iPad Pro + Pencil lover. And
I'm very tired of carrying both a MacBook for the full OS + keyboard, PLUS a whole separate tablet for sketching, illustration, annotation, etc.
I think there was a modifier key + click and drag to simulate a pinch, so at least some gestures can be simulated. But since one is developing for the mobile device, it makes sense to tether the device and send builds to test out on the target device itself.
From past experience (not for quite a few years, but some at least): oh heck no. I loved simulating locally, but the deeply limited touch handling made it infeasible a lot of the time. The dev -> install -> run -> dev cycle is dramatically faster in the emulator, and you don't need to have multiple devices nearby if you're building custom UI for each size.
Macbook Pros attempt to cater to every slim use case simultaneously, with the idea that if they can handle a use case someone else can’t, they’ll sell another laptop.
I agree that a piano demo seems like a bad idea. But damn if I don’t daily long for the iOS 6 days where buttons and tappable areas were well defined and there were some fun and whimsy to the interface. As compared to the flat unintuitive desolate landscape of recent years. I would rather have a big animated cassette deck in my music player than the current overly polished full page advert I get every time I open the app. At least Forstall and his team cared enough.
I believe that user interfaces have just got appallingly bad. Compare Windows 3.11 which came with a paper manual and had about 5 pages detailing scrollbars, buttons, menus and window controls. That is all you had to learn, and you could use the system through consistent usability.
This seems to have been universally abandoned so that user interface elements are hidden (scrollbars), buttons don't look like buttons, and you can "scroll" or move window contents even if there's nowhere for them to scroll to (like on an iPhone where you can "scroll" past the boundary and it comes back. It makes everything an exercise in frustration as you have to grope the user interface to discover if anything is usable.
I believe this is why some of the older generation have such difficulty with basic features on phones and tablets - nothing behaves consistently, and nothing looks like you should be able to press it.
As a longtime laptop musician, using the QWERTY keyboard to play notes in the absence of a MIDI controller is infinitely preferable to this goofy demo.
Apple better have some better use examples than THAT.
Reminds me of that stupid DJ demo of the touch bar. Did any DJ actually use that in practice? Well if they do they better hold on to their now vintage MacBook.
Yeah, piano playing was a bad example. There’s a reason MIDI keyboard manufacturers try so hard to replicate the weighted feel of a real piano. Your fingers need feedback.
An interesting aside here, there is no one, true style of action for keyboard instruments and no one, true type of feedback.
I have a good 88-note hammer action stage piano. I love it for acoustic pianos and it's OK for electric pianos (Rhodes, Wurlitzer) but it sucks a bit for organ, clavinet and synths.
I've seen people play Hammond really, really well on a weighted keybeds but it's much easier to just plug a second cheap(er) synth action controller in for organ parts.
I've never met anyone that really liked playing a tablet screen though. :P
How is it convenient? A pro workflow is not holding your arm with the finger up in the air trying to avoid accidental taps, when you can get a small 25 key akai midi controller for a 100 usd. And it's easy to carry around and comes with velocity-sensitive keys and pressure-sensitive pads.
Similar 75% price cut trajectory as GPT-5.6 Sol -> GPT-6.1 Sol. Overall, Terra is still half the price of Sol.
But like the original release of Terra, it's still a weird middle child between the small-model-that-got-substantially-better (Luna) and the large-model-that-got-significiantly-cheaper (Sol). The TerminalBench 4.0 cost-per-task results are a bit confusing and I'm not sure why you would pick Terra over Sol in that case. Overall this release seems like a flex on Sonnet 5.5: a model which is having its own middle child problems between the godly Opus 5.5 and the surprisingly capable Haiku 5.5.
Due the the size of the default system prompt you get over 100k really in easy in Claude, and because that is fixed 200k is really almost 3x more room to work with, not 2x.
200k also used to be the limit for Sonnet for a long time.
If Haiku 5.5 also had 200k, it'd actually be usable for us. As it is, it's Luna or Sol.
On purpose? I was using qwen3.8-27b the other day and very quickly ended up burning through half of the context (it has 128k total) pretty quickly. Not because of my directions, but simply because the codebase was very large.
256k seems sweet spot. Even large codebase can be handled you just has to map out areas for agent. Anything past 256k usually exception than norm to be.
At least this my discovery when I did evals on my reviewers skills. 10k LOC is high end for high effort or around 150k context using opus. You can push towards 15k but that’s on max effort with about 400-500k ctx.
Fast models like haiku do not even need 1M ctx in my view
Yeah codebase structure is very important. Admittedly, by large I mean average size but very poorly structured with a spaghetti (and sometimes cyclical) import graph.
Artificial Analysis still has MiMo 2.6 Pro setting the Pareto frontier where it matters: it's as cheap as Terra High/Sol Low while being smarter than Terra Xhigh. Sol Medium beats MiMo in smarts but is priced a lot higher. Meanwhile Terra Max is dominated by Sol High in both price-per-task and smarts.
I recently tried MiMo 2.6 Pro on a simple app and was positively impressed (modular code with sensible tests). I then tried it on a more difficult problem (which Opus 5.5 solved in minutes) and it looped for 12 hours without making any apparent progress before I finally pulled the plug.
It is having improvements in other fields than coding while being more cost efficient.
If all I want is coding at a cheaper price, then deepseek is a good model for that.
I am more than okay to have a bifurcation of models good for coding and models good for everything other than coding. I strangely imagined it many years ago.
Everyone is suggesting OpenAI to release a model close to opus 5.5 yet it's releasing this. On one hand we can imagine that maybe their training cycle wasn't able to produce this whereas gpt 6.1 terra was made.
But on the other hand, I also now wish to see an OpenAI model which is good for just pure coding, like an absolute beast at it while being cheap (so the opposite of terra)
ironically, one of the few appreciations that I might give to OpenAI if they pushed this model to prod and are facing some pushback as developers like us want/only see coding models. They could've benchmaxed it and it would've been slightly irrelevant. I'd much rather see reality.
At the very least, this new model won't make things more expensive for anybody and would only make use-cases cheaper depending on what they are, so it's a fine deal for me.
I also wonder maybe this is why open source labs aren't targeting this space and why the only focus of AI labs is coding nowadays. Everything else just doesn't seem so worth it from investor, even end user, training perspective.
Maybe this is the reason why we have gotten almost an exponential growth in coding capabilities as that became the most (if not only) capability that truly really matters.
To not wish for discussion to go to a tangent, nonetheless this could also explain why our work as a software engineer is going through a slight crisis of meaning as these AI models have gotten impressive at coding.
As an (aspiring) developer (currently in 1st year college), I don't wish for AI to replace coding even more for me or to push it even more or to suck the joy out of it even further. So I am more than happy for these models to stay as they are and rather make advancements elseplace. Ironic because it is us developers who use AI as well and we want cheaper workflows and so we would also want models to be cheaper and better at coding. There are some fine lines of contradictions and irony in all of it. I'd much rather also have open-weight models of such terra-like models but I imagine that because of the reasons that I mentioned above, we are most likely gonna see it the most focused on coding but still open-weights is pretty close.
The more choices the better (though choice-fatigue is a real thing), I hope I was able to explain some of the nuance I feel about this particular launch.
For technical work, I’m inclined to agree. I’m a fan of Luna and Sol, and Astra has its place, but I’ve always found Terra an unhappy medium.
If Luna is not good enough for a task, then the marginal improvement of Terra tends not to be worthwhile. If it is worthwhile, then the jump to Sol tends to also be worthwhile, so I tend not to use Terra.
While it does offer some meaningful improvements over 6-Luna for a few workflows, I suspect most of those benefits stem from 6.1-vs-6, as opposed to Terra-vs-Luna.
For me, I used Zed Agent with Codex as it has native integration with Zed's ACP so it's pretty convenient.
However, Anthropic is pickier about the harnesses you can use with Claude Code and I noticed discrepancies when using Zed's External Agent feature with Claude Code, so I just use the Claude app. It's fine.
reply