GPT-6.1 Terra
openai.com
openai.com
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.
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.
Destroyed by Anthropics crazy decision to price it x5 after 100k context.
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
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.
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.
Looking at newer models, Terra 6.1 is only a hair behind Sol 6.1, hopefully making it a useful Sol replacement for slightly easier tasks.
Luna 6.1 is fine with it, but the results aren’t good enough. I’m going to try Terra 6.1 (from the post, I should expect about the same results). And hope it doesn’t complain as much about these non existent cyber security risks that for some reason it keeps raising.
P.S. I also tried Deepseek Flash 4.1, but the results simply didn’t cut it.
P.P.S. This all investigation was a major eye opener on why we need to run local models or be forever tied to some artificial limits to AI for the masses.
No, they can't, considering Terra never existed for 6.0. It's a mess, and no self-respecting user would care for it.
Also, you can just…wait a little longer until all the models for the version number you want are available? You don’t lose anything, so how is not “sane” to do so?
First we had:
> Luna (Small), Terra (Medium), Sol (Big).
Which was fairly smooth after you got used to their planets.
But then they added Astra as the "Bigger," which was mostly still self-coherent:
> Luna (Small), Terra (Medium), Sol (Big), Astra (Bigger)
But then things go confusing, they released Sol-6.1 which sometimes is superior to Astra, and discontinued Terra which made it look like this:
> Luna (Small), Sol-6.1 == Astra
But with Sol-6.1 costing the old Terra-6.0 price, while, as I said performing as good as the old "Bigger" model in many non-3D/UI situations.
Now they're re-introduced Terra, as a 6.1 release, making this even more confusing.
Did Microsoft's chief branding officer move over to OpenAI?
But minor credit to Anthropic, at least they've mostly stuck to their four model names (Haiku, Sonnet, Opus, and Fable). You mostly know where a model is MEANT to be positioned from the name alone. OpenAI cannot keep it consistent from release to release.
But, yes, currently Opus Vs. Fable has the same issue as Sol-6.1 Vs. Astra.
Hmm? They've been using the Luna/Terra/Sol/Astra scale for a few versions now and as far as I can tell it maps directly onto Haiku/Sonnet/Opus/Fable, in terms of the kind of thing it's meant to convey (if not in terms of the scale of differences).
Both companies have screwed up their naming royally. Maybe that's intentional - I'm sure it doesn't hurt Anthropic to have people using Fable a bit longer than they need to and racking up a huge bill in the process.
It's like iPhone 17 being faster than iPhone 16 Pro. Because it's newer and benefits from advances in chip architecture, despite being a lower model tier. Sometimes different iPhone models are released at different times.
I think the names are more of a reference to price. Luna is always cheaper than terra which is cheaper than Sol which is cheaper than Astra (for the same model number).
Per token, yes. But per task it's not even necessarily the case depending on how much thinking is needed.
But the model names are linear by price.
Yes, a more modern smaller model can beat an older bigger model at some tasks, that's what can be confusing.
“Test them all”
The lineup of models is still the same.
There are four models: luna, terra, sol, and astra.
Pricing increases for each level. This new 6.1 Terra is 50% cheaper than Sol 6.1.
They are lowering prices for newer releases, so yes if you compare Sol 6.1 to 5.6 Terra it won't look right, but why are you doing that? Look at the same release version for comparable numbers.
Each model improves in successive releases but they maintain pricing and capability coherency within a release version.
Yes, Sol 6.1 is supposed to rival Astra for some work, which is what you'd expect from a later versioned release model against a prior release bigger model.
Assuming they release Astra 6.1 at some point, I expect it'll be better than Sol 6.1, at a higher price.
I don't understand what people are asking for here? No reductions in pricing? Different model names on every version release? Not releasing any models for a version number until they're all ready?
- Model name
- Version number
- Effort level
Given the operations of the model are opaque, can you see how someone (most people?) might be confused?
After that you just have model "strength" and version number. But of course strength is only comparable at equal version number and the two axes are almost parallel.
Anthropic also does this, perhaps to a lesser extent.
I see a ton of criticism and not a single person coming up with anything better, unless it's to take away options, like removing effort level.
What about buying Intel CPUs? A car? Electrical components?
If you're not just using this at a consumer level to ask about cooking recipes (and if that's your use case, you really don't care about API pricing, or any of this, and can just use the default), and are using it professionally, then this seems like a low bar of effort to understand what you're buying.
I do grant you the relationship between model and effort level has an unclear relationship, but there are also tradeoffs in most complex technical products that are not always clear; this is exacerbated by the fact that this product is stochastic.
Why are engineers using this as a professional tool expecting this to be bought like a consumer gadget and then blaming the maker?
Why is it ok for engineers to have confusing names and ways to interact with something?
How am I supposed to decide between these options: https://imgur.com/a/bXHWowL
How am I supposed to keep up with the changes every few months?
Are you presenting this as evidence that OpenAI's naming scheme is bad?
If you don't want to spend the time to know the difference between all these different suppliers and their models, then why are you using whatever this program is that gives you so many choices, most of which are bad? Why is this OpenAI's fault?
No worries, you can always ask a AI.
They should have a guide saying "Terra is best at technical documentation and communication, Astra is best at large software projects and research, Sol is good for development" or something like that.
This morning I asked GPT which models and settings I should use to inspect my VM dashboard with computer control. It said: - GPT-6 Astra at Medium - GPT-6.1 Sol at Medium (alt) - use Luna only when the route was already known.
But I had already completed the task twice with Luna and it was 5.6! It navigated through a few pages of my dashboard, pulled up the 30-day usage chart, analyzed my CPU and RAM, and explained why I didn’t need an upgrade. I’d give Luna 5.6 an A on this task.
Maybe OpenAI should give us a “Tiptronic” selector: recommend the model and reasoning level based on the prompt, while still letting users override it.
What are you talking about? If it's for some slop you vibe check it. If it's for a serious effort you write an eval. Same as it has been for the last 3 years.
Let's say I have some task that Luna 6.1 cannot do, but Terra 6.1 can. So I build my product around that. In a month when Luna 6.2 is released, then I have to figure out whether it's worth to evaluate and swap or just wait for Terra 6.2.
Waiting is the way to solve this. And despite the outcome being the same, making me feel like I'm being forced to choose to wait is a worse consumer experience.
They released gpt-6-sol (alongside astra and luna but no terra) and 7 days later released gpt-6.1-sol making gpt-6-sol basically DOA before any real adoption. Everyone knows it wasn’t just better RL/post-training, it was a completely different model underneath. Now they randomly reintroduce terra.
I’m sorry, but if you’re not even a tiny bit befuddled by this then you must be taking whatever OpenAI is having.
If you scale a model down, or switch out attention with linear attention, or introduce KV-compression, etc., you don't really know how it will affect model performance before you've trained the model.
And once you've spent millions training it, you might as well make it available if it's pareto optimal for some tasks. It also gives you more insight seeing how people use it in practice.
You are also rolling the training of your models by size and the lat est small model can sometimes beat an older model.
For a while we thought you just train the big model first then distill smaller models from it but since we've learnt the importance of on policy RL (basically the way a small model figures out a problem is very diff from a big model so it's best to RL optimize each model distinctly)
And some variants (e.g., the medium one) being skipped or delayed for a certain generation doesn't really change the overall size branding.
It's nice that they have Terra which seems to fill a void that's been needed since Sol 6 focused on cost (at the expense of quality), Sol 6.1 got a huge boost in quality, but became unusably slow unless you paid Astra prices - at which point, why not just use Astra?
Now at least we have a viable other option again.
In general, LLM performance is incredibly multidimensional. There's no guarantee of a fixed ordering where the model that is best at math is also the model that's best at jokes. Despite that complexity, much of the variation in performance is explained by just a single variable: model size. In general, models that are bigger will (a) be more performant / reliable, (b) cost more per token, and (c) have better recall of esoteric facts. As you note, our models are ordered:
Luna < Terra < Sol < Astra
Even today, I'd recommend this as a general starting point for picking models. If you care about quality only, pick Astra. If you're a bit price sensitive, pick Sol. If you're very price sensitive, pick Terra or Luna.
Of course, model size doesn't capture all variation. Models that are trained later might have new data or new techniques or new architectures that earlier models lack. You can see this with models from all providers - Sonnet 5.5 outperforms Fable 5.1 in ways, and Gemini 3.8 Flash outperforms 3.1 Pro in ways.
In the case of GPT-6.1 Terra, I'd say Terra is probably better than Astra and Sol at technical communication, similar on vision and computer use (but much faster!), and a bit worse on most other things. For API developers, it's a nice midrange option if you want something cheaper than Sol but a lot smarter than Luna (especially for vision / computer use / communication). For subscribers, it's not a bombshell release - if you're hitting usage limits, maybe it helps you go a bit further, and if you want clearer communication or faster computer use, it can be a nice option to select.
If you have suggestions on how to name our products better, I'm all ears. Our intention is to make things clear and simple, even if we miss the mark sometimes.
Unfortunately, a lot of the complexity is somewhat irreducible and due to to the actual complexity of the products. Scaling model size is different from scaling reasoning effort which is different from scaling multi-agent count. So if we just swept all of those under a single name of "more IQ", it would make the naming simpler but the experience much worse, as important details would be obfuscated from end users.
> Despite that complexity, much of the variation in performance is explained by just a single variable: model size.
please release model size :p
For some reason, this is just hilarious to me. Idk if you're being serious or if this is passive aggressive but the thought of OAI actually taking advice from a random HNer and basing their naming scheme off that is peak.
Where else would you go? I'm honestly scratching my head thinking of someplace that wouldn't run you out on a rail, but that would give serious feedback. Kernel mailing lists are out, maybe there's something like that which I can't remember off the top of my head, but...
Also, nomenclature is incredibly hard. There are only two hard problems in computer science: cache expiration and naming things.
The primary users of any API are the designers and engineers of the applications that use it. How many laypeople know what Postgresql is? How many of the applications they use run on it?
- Understanding how effort and class interact (Is a high effort terra above a low effort Sol? Are larger models mostly always better regardless of effort?)
- understanding when (if ever) a new model tends to be better than an older but higher class one. Is an old astra better than a new terra?
If increasing effort and increasing model class produce improved results in a different way, I miss making it explicit in which way. In other words: What kind of issue warrants going up in effort? what kind of issue warrants going up in model class? What kind of issue is unlikely to be solved by either?
Oh, easy:
Artificial Analysis has their own curves which are pretty similar.
You can literally see this in the charts from the article
That could solve the 'more IQ' problem, because you'd actually have different dimensions (e.g. for 'raw intelligence', technical writing, emotional understanding, etc) that you could use to communicate the differences in models better.
(Of course people would instantly start complaining that the benchmarks numbers are probably fake etc, but it is what it is.)
I.e. are all version numbers independent and Model X 7.5 is not necessarily of the same generation or cut off as Model Y 7.5? Or do all of your 6.1 models have something in common - cut off data, generation, baseline technology, approach, architecture, or something else?
Thanks! :)
Back in the day we used numbers to also sometimes indicate size, which could make it extra confusing: e.g., we launched GPT-4.1 (an incremental update on GPT-4) after GPT-4.5 (a new model that was a big jump over GPT-4).
Terra and Luna are dirt cheap and good enough for 50% (or more) of the typical junior dev's workload, and great for routine work. Sol and Astra have so much written about them at this point that it should be obvious what those are good at. The large/gigantic models for things that need full intelligence and slightly more branching thinking.
Luna < Terra < Sol < Astra, fine. But does Terra 6.1 beat Sol 6? Does Fable 5.1 beat Opus 5.5? If I build something around this Terra 6.1 release, then when Luna 6.2 is released in a month, should I swap?
If they were called iPhone 14 Luna and iPhone 16 Sol then this is less clear.
Or on the flipside, if the models were called GPT 6.1 Mini and GPT 6.1 Max, nobody would be arguing about the names being hard to grok. The words 'mini' or 'small' are more descriptive than 'Luna' or 'Terra'.
If you look back the naming problem started around the point that OpenAI started doing reasoning models (o1, 4o) where they had to describe this fork in model behavior that let a smaller model think longer and be smarter.
You can come up with naming schemes that don't clash, but there's something to be said for this approach where the thinking level is like you describe and the model name requires you to learn that Astra is the largest model.
* Luna (Moon)
* Terra (Earth)
* Sol (Sun)
* Astra (Galaxy/Universe)
The relative sizing seems dead simple.
You're right! The relative sizing of models is dead simple. This doesn't help when there are improvements between release numbers, and models are released one at a time instead of as a family.
When Sol 6.4 comes out and the latest Astra is still 6.1, which do you use?
Or I might try both and pick whichever I prefer for a given task...
1. Even more dead simple if it just used Moon instead of Luna, Sun instead of Sol, etc. - one less quick translation in the head.
2. Or even more dead simple if it just used "Small", "Large", etc. - removing another quick translation in the head. (On the other hand, all these suggestions are English - maybe Luna, etc. is more understandable for non-English speakers?)
Is a sunburst supposed to be bigger than a flare? I mean you get solar flares too? Or is a solar flare bigger than a sun 'burst' which usually just means some sun coming through clouds.
I wish
Sure, in a hypothetical world where the latest Max phone is two generations older than the latest Mini phone. In which case now you have to evaluate not only screen size but ALSO specs and component age and software support etc. How badly do you need the bigger screen? Etc
If you are comparing the iPhone 16 Max to the iPhone 16 Mini, everything is a lot easier.
I'll just stick to mostly Sol on low/medium/high and start experimenting with Terra tomorrow onward to see what it does noticeably better than Luna/Deepseek flash.
Any pareto frontier model, especially from the leading labs is guarnateed to be good enough for whatever your expectations are at this point. A quick few day test should be enough for anyone to make their own conclusions.
> Run your own evals then if you have this much of a need to "correctly" choose.
No. OpenAI want my money, it's on them to communicate properly.
But here we have LLMs, wherein: It depends. The smaller models can in fact outperform the bigger ones, and get some stuff done cheaper, better, and faster.
If the question is "Which one is best?" then the answer is "It depends on what the fuck you're doing."
(If that answer is found to be unfulfilling, then let it serve as an indication that a person may be better-served by improving their questions.)
If you want to go optimize your mouth hygiene it is actually more important to use dental bristles (flossing is not as good), because most of the work is done by the mechanical cleaning action of the tooth brush and you need to reach the places where the tooth brush cant get into.
Once you have done that, the next step is alcohol-free xylitol based mouth wash. Caries bacteria have trouble digesting xylitol so this reduces the presence of them in your mouth.
Only after you have done that, are you ready for something better than fluoride. Fluoride is a remineralization agent. Your saliva also contains calcium that helps your teeth remineralize. So the next thing you need is a more potent remineralization agent. Basically your tooth paste acts as a more concentrated saliva. Because of that, it is actually better to use e.g. fluoride in the morning and a different agent in the evening.
The classic options are bioglass (novamin) or hydroxyapatite. The latter is also available as mouth wash.
I understand how it can be confusing, but I don't think that this is exactly a problem.
Luna through Astra offer qualitative descriptions of the model's size. Presumably OpenAI trains its biggest model (Astra-ish, maybe even larger) first, then distills it to smaller sizes.
Verison numbers reflect something about the architecture or training dataset. Again, they keep these details close to their chest, but we should expect that weight-for-weight higher versions should beat lower versions.
This offers us two axes of 'better', but these axes are also not directly comparable. An earlier but bigger model may or may not beat a newer but smaller model. This is where 'spiky' intelligence comes in.
This is also where calls to 'run your own evaluation' really do start to matter. If you care enough about this performance[1], then you also have specific problems that you care about. No generic benchmark can fully answer that question, hence the need for a specific evaluation.
[1] — And you don't have to! A casual user is probably served reasonably well by anything put out in the past six months. I think Microsoft is still serving GPT-4o under some Copilot branding.
Whether they train the biggest model first or not is not relevant; what matters is what they release. Right now we've got 6.1 Sol and 6.0 Astra. Over at Anthropic they've got Opus 5.5 and Fable 5.1. The existence of a hypothetical Fable 5.5 that I'm not allowed to use doesn't help me.
I wouldn't necessarily say that. Atmospheric mass may be obvious to us, but I suspect it's not nearly as universal knowledge as one may assume.
And I actually assumed at first that they were based on distance from us (terra < luna < sol), so I kept confusing terra and luna.
A favorite…
Prompt: “Tell a joke about mother in laws”
Gemma2 response: “What’s the best kind of mother in law? A dead one.” 0_o
We're probably headed to further specialized agents where a single ordering or even model size won't mean a lot.
What's confusing to me in addition to the naming is that there are no published guidelines I can use, as far as I can see. Put yourself in the user's shoes. On what basis should one pick a model for a given task?
I can understand smaller models being dumber and faster. That aligns well with scenarios where you are willing to compromise; e.g. you're classifying content or asking a model to format text into a table.
But when you want "maximum intelligence, cost and time be damned", there's the implication that bigger models are smarter on a linear scale, and every provider's model release naming is currently conflicting with that expectation.
This isn't just a naming problem, but a communication problem. If I knew that Astra was the best at coding but a bit bad at writing, and Sol was almost as smart but better at technical writing, then that's the kind of thing I expect to be documented.
One possible way: A simple matrix with horizontal columns like "writing", "coding", "math", rows are the models, and each box is a number of stars from 1-5.
Instead, this sort of thing becomes lore, shared by word of mouth.
I don't know, to be honest. Technical documentation is such a multidimensional skill, and I also don't know your budget. I'd imagine Terra might have the most readable prose, but Astra might be more likely to catch bugs or mistakes to highlight. If the documentation is about a more esoteric language or situation, Astra might have a greater edge than if it's a common language or situation. If you have a long list of documentation guidelines, that might also give more of an edge to Astra. I wouldn't want to make a recommendation without doing a lot of testing, and I haven't done a lot of testing.
But to assess whether my preference on my problem is going to generalize across other people, other programming languages, other documentation types/formats, etc., and then analyze statistical significance is a thornier question that I don't have the data to answer. I'd rather make no recommendation than a weakly informed recommendation.
That's what users currently end up having to do on their own. Or they just give up in confusion, which presents in your site analytics as either bounce, buy later or buy less.
Why not have a model comparison site? Visitors choose from a dozen-ish broad categories of work (eg technical documentation, market communications, etc), then pick from a dozen sample prompts across typical tasks in that category, and see the model responses with avg time and approximate cost. Especially helpful would be having a toggle overlay with annotation highlighting relevant differences, then if they have (or create) an account, they can edit the annotations and add their own with a sharing link.
If my job was to actually help my co-workers and friends make better informed choices they feel more confident in, I'd be doing something like that. The extra credit version would help users narrow down to two models and then let them convert to a 'Test Drive' trial - same workflow but from their own prompts, side-by-side A/B response comparisons with annotation and sharing link.
As you said, the answer to "Which model is best for me?" is actually complicated, conditional and inevitably a bit confusing. When that's the reality, the right pattern isn't 'Communicate to Convert', it's 'Build credibility and confidence with maximum clarity and genuine helpfulness'.
People vote on pairs of model responses and they aggregate everyone's votes into a leaderboard.
One thing to be aware of that the distribution of prompts on arena.ai can be pretty different than real-world usage. The people who spend time on that website casting votes tend to be more interested in entertainment than professional productivity. And turn-based voting can punish useful behavior like asking clarifying questions instead of guessing at assumptions.
It’s overkill most of the time, but frankly all these models are “cheap”, and I’d rather waste compute than brain cycles trying to discern which dials are best for each individual as-hoc query.
Every model I've tried is just abysmal at this, for what it's worth.
If any do it well please let me know and I'll give it a spin.
I recently experimented with some image generation models and I was surprised at how much context/prompting is needed to get something that doesn't look Lovecraftian. There was seemingly no downside to adding more and more detail up to the limit. Prose sits between code and images in terms of fuzziness and I wonder if promptmaxxing is the winning strategy for technical documentation.
I've been thinking that fine-tuning a model to write differently shouldn't be that hard (tedious, perhaps) and if I had the resources I would consider doing it myself.
Create a realistic eval suite for your use case using your actual data that you manually label, with easier, more complex, rarer samples, and run each model you are considering against it and decide based on the usual confusion matrix, cost, latency, or whatever metrics matter to you. There's really no other way to know – and you might be surprised that cheaper models can often perform your task(s) relatively well at a fraction of the cost.
I’d imagine there’s only a tiny percentage of users that can or want to do this.
but if you want to use some LLM agent for anything commercial and you want to get the best and at the same time you don't want to overpay the easiest and the surest would be to create your own benchmark for your specific use case.
I do not use agents to write on behalf of myself. I write my won technical write-ups and my own slide presentations. A lot of people don't; and it shows. There's much that agents can do better than humans now, but not everything.
It's high-level (and doesn't break out coding/writing etc), but https://developers.openai.com/api/docs/models says:
> If you're not sure where to start, use GPT-6 Astra, our flagship model for complex reasoning and coding. Choose GPT-6.1 Sol to balance intelligence and cost, or GPT-6 Luna for cost-sensitive, high-volume workloads.
That page also seems to be updated after each new release.
Thanks. I think the very response you wrote here is a better form of marketing for the software engineers than what openAI puts out there. When everything is groundbreaking and everything tops charts, nothing is.
I suspect a lot of target demographic (developers) prefer simpler and honest communication about not just the models high points but low points too. With the linked Terra post, if I saw 1 chart that clearly highlighted the usecases for which Terra is _worse_ than higher tier models, I would be more convinced overall.
Everything trying to occupy the top spot in my brain's limited "top model" representation space is not helpful.
The website linked starts out with:
> We’re introducing GPT-6.1 Terra, a fast, efficient model that combines detailed visual reasoning and precise computer use alongside clearer, more natural writing. A significant step up from GPT‑5.6 Terra, it excels at the communication and computer-based tasks that fill the workday, from drafting proposals and interpreting technical drawings to filling out forms and editing videos.
If you go father down, it has graphs and details with specifics of this, including:
> We measured GPT-6.1 Terra’s computer-use capabilities... At medium reasoning effort, GPT-6.1 Terra scores 15 percentage points higher than GPT-5.6 Terra at about one-third the cost. It even performs on par with GPT-6 Astra at only one-tenth its price per token.
and
> On MMMU-Pro... It even nearly matches the scores of our most capable frontier model, GPT-6 Astra, and GPT-6.1 Sol at much lower cost.
It’s probably just unavoidable. Like when Apple release a new MacBook Air first and it is (in some ways) faster than the old MacBook Pro.
I think the planets thing is a bit confusing though. I hadn’t really thought about it being based on the size of the planet. I assumed it was based on how far away they were with terra therefore being the base one.
Instead, we got Opus 5.5, Sonnet 5.5 and Haiku 5.5 released weeks apart, and so are Sol 6.1 and Terra 6.1?
all training runs have to at some level share the compute capacity ; while they have more than enough to run many large runs concurrently, they don’t have enough compute to run everything concurrently.
Personally, I pretty much exclusively use claude for coding work (sorry!). Our application is running almost 100% on OpenAI API (except for some fallback models). It is the best API on the market. The models are excellent particularly for any application that is cost sensitive.
it's almost like ya'll let the AI name it ;)
I don't expect the vast majority of your users care. I would guess they want a rough proxy for "cost" and will assume cost=>capability are roughly linked in a linear fashion.
it is very possible that the users weighted by spend do, and I'm therefore simply wrong. But, 2c
Have a model alias called "math", another "jokes". Quality labels/date suffixes to these obvious labels. Repoint as needed.
Something for people who don't want to play the cute "which part of the solar system is trending this week" game.
Even "writing" as a single concept is very multidimensional. There's fiction writing, nonfiction for experts, nonfiction for learners, deep research, quick lookup Q&A, copyediting, writing in documents / artifacts, instruction following about style, and default quirks (e.g., em dashes, negation, loving the word "tapestry"). Some models are great writers but less diverse, meaning people prefer the first story they read but disprefer the tenth. How does one label such a model?
But this is exactly what we have! Except it's worse, because the model usage recommendation isn't in the API, it's in some random person's X advice that's no longer current this week.
I understand AI companies wash their hands of this problem because it's hard, and they themselves barely understand the true capability matrix of the models they're selling.
I just don't think some effort to categorize models and expose that to users would be messier than what we have -- model selection driven by influencer voodoo bandwagons and analogies to literal astrology.
> If you have suggestions on how to name our products better, I'm all ears. Our intention is to make things clear and simple, even if we miss the mark sometimes.
How about naming the models:
"Small", "Medium", "Large" and "Very Large"
Or refer to them by parameter count like other companies (e.g. this is 6.1 5T)
If I have a clothing store, and rather than name the clothes sizes Small, Medium and Large I decide to call them Urgon, Planfar and Fangor I shouldn't be surprised if people think my size naming is confusing or silly. And IMO it's not a great explanation to say "oh it's easy, if you are skinny then you should get Urgon, but then if you are a bit more of a medium build you should get a Planfar, but if you want something a bit baggier than Fangor will really work out for you"
As a point of comparison, GPT-3 was launched in a series of small, medium, large, and XL. Back then, XL meant a mere 1.3 billion parameters!
You wouldn't - you would just put the version before it - 6.1 Small, 6.0 Medium etc. Not that complex.
Think of an iPhone 18 Pro - No need for 'iPhone 18 Even Better Super Pro' because the 'Pro' is relative with the version/generation.
Hire an actual marketing person instead of letting engineers / ML researchers do things they're clearly terrible at ?
This seems like one of those cases, where CalVer would be appropriate, and you likely have ideas what you aim for while training, so technical model names (ids?) could include aliases to most things you said. Like 2026q4-xlarge-researcher, 2026q4-medium-humanitarian, 2026q4-medium-coder, 2027q1-small-analyst etc.
Well you definitely missed that mark extremely badly here.
I can’t tell you the difference between the 10 different versions of the current iPhone and I can’t you the difference between the 5-6 different models put out by the Anthropics/OpenAIs of the world. At this point I just try to track the flagship and then the cheapest one or worst case I just use the flagship.
I’m sure internally there’s data showing people want choice and options but decision fatigue is of course very real.
With an optional flash mode.
Then as you release models, you pick the model and thinking level that you believe is pareto optimal for cost/performance.
So this auto updates each release, like Terra was probably off this list entirely, but maybe this would appear now in Auto medium or something like that.
I know from looking at charts that each model seems to have sweet spots (like I always use this model in this thinking level and if i want to go cheaper i go here). If you picked those for us, it would be interesting. I do think you should keep the model families for those wanting more consistent experiences, but for many devs they may want to just pick auto something and have it update over time. I think nice if it actually just said in parenthesis what it was, something like: Auto Medium (Terra 6.1 High)
I am not sure if everyone would agree here though. I'm obivously advocating for more naming schemes, and i understand the irony.
xlow = Luna Low
low = Luna xhigh
medium = Terra medium
high = Sol medium
xhigh = Sol xhigh
Max = Astra something?
Tiny/2026-11. Medium/2026-6. Big/2026-10. XL/2026-01. It may not be sexy, but people know exactly what it is and what to use.
Of course we have far too many benchmarks (of questionable utility), but it would be helpful to get OpenAI's take on where it slots in, and how the landscape changes as these get rolled out.
GPT 6.1 X > GPT 6.0 X GPT Y Astra > GPT Y Sol > GPT Y Terra > GPT Y Luna
This order is very important.
I'm just worried the system design won't be as good if I choose a lesser model so to speak.
Edit: typos/grammar
You certainly know that it happened not by accident that traditional computers are producing reliable results, a lot of work went into this.
Computers must produce reliable results, they are simply not useful if not.
People will understand that at some point, but a lot of damage already has been done.
You have Luna, Terra, Sol and Astra, each with 4+ reasoning levels. That's ~20 options to choose from.
On a repetitive task like customer support or something, where you have decent benchmarks, you can plot cost versus performance, but you can't really do that on coding.
Terra 2026.2
Despite Opus 5.5 being newer and "better" if you read benchmarks, I've found Fable to be better for some things (subjective stuff)
>If you have suggestions on how to name our products better......
Too late for that. I hope you guys can at least stick to these naming scheme for at least another 2-3 years going forward. Not everyone follows AI / LLM intensely, and it seems there is a new naming changes every year or so.
At least conceptually OpenAI's naming is easier to understand than Claude.
It terms of capability, it would be great to make it clear somewhere that this is 6.1 Model "+" technical communication, vision and computer use improvements. So that at least we know what we should try with these new release and choose a better model for certain workload.
Right now, other than pricing, it is confusing as hell. And in hindsight I wonder if there are fatigue in choosing one people just go with Open Model and stick with it if it is good enough.
<version>-<size>
Where <version>: 6.0, 6.1, etc., and <size>: xs | sm | md | lg
E.g. 6.1 Terra = 6.1-sm
If you eventually make specializations like "math" and "jokes" then I would call it something like 6.1-jokes-sm.
Versions are good for backend but when communicating I want 2 or 3 levels of detail.
Class_Generation_Modifier
Astra_8_Fast
That's it.
This means: Class - Tells me what it can do, how well. Generation X end of life is ~18 months. Modifier: Variation of options.
And the modifier can be dropped if you add more classes.
I'm not the one you were talking with but catching you is godsend since I just ran into a lot of confusion on the naming myself. I think some of it has to do with how they are labeled in the app.
I think simple numbers in naming is great, even 6.1 is fine bc I understand that if I want the latest model, I just need to look for the biggest number. Where I got confused is in the "work" tab, I saw 'Astra' is set as the default. I see the label for it is:
"Frontier intelligence for your most demanding tasks."
To me that reads "this is the 'best' model." Then I see Sol-6.1 which is labeled:
"For coding and every day work."
Then Terra-6.1 is:
"Terra 6.1 Codex Model" (tbh this one might be the most confusing, I'm aware of codex the product but Sol is the one labeled good at coding)
From my POV as a basic plan-based user wanting to do something like...
> design a robust event-driven system, here's the details [...]
I honestly have no idea which one I should use.
Would Astra-6 produce a better design than Sol-6.1? Is the '.1' a big step up from Sol-6? (I'm guessing no but I truly have no idea). I would assume the answer is Astra-6 would produce the best one, even though the labels don't imply it is necessarily geared towards coding?
But tbh the "default" seems to always work for me... and after reading your comment, typing this one out, being forced to think about it... it doesn't seem too bad given your challenges in communicating the purposes of each model and wide audience of consumers. I'm also sure anyone could find a million problems with decisions I've made in software I've built so I try to not be too harsh :)
Maybe stop using version numbers then and just name models like you'd name children.
Medium v10 vs Big v2 makes it look weird and confusing. Especially when there are a sea of influencers and twitters with vibe assessments who will tell you new models best olds models.
Can’t you just start at the top or bottom.
Release new big for paying customers -> release new medium for paying customers -> release new small for everyone
Rinse and repeat. We’d all be much happier
Also, I think the naming is great. Like you said, Anthropic faces the same issues you do. Opus is much better at Fable in a lot of stuff without being quite as huge or as smart.
Congrats on the launch.
Is this really so/could you give a rough example please? I've not used a lot of Claude or ChatGPT, but 3.1 Pro always feels most grounded and intelligent. 3.8 Flash (in my very limited testing) isn't bad, but it feels less nuanced.
(Fwiw I don't use it for code, but ask maths and real world logic based questions.)
Ted,
I'm assuming you're not in the OAI marketing dept because leading with a simple thumbnail like "Luna < Terra < Sol < Astra" is helpful framing and dropping next-level bombs like:
> "Terra is probably better than Astra and Sol at technical communication, similar on vision and computer use (but much faster!), and a bit worse on most other things. For API developers, it's a nice midrange option if you want something cheaper than Sol but a lot smarter than Luna (especially for vision / computer use / communication). For subscribers, it's not a bombshell release - if you're hitting usage limits, maybe it helps you go a bit further, and if you want clearer communication or faster computer use, it can be a nice option to select."
Is just far too clear and useful! I'm kidding... but it's funny because it's a little bit true.
> Our intention is to make things clear and simple
Your post is the single best product marketing communication I've ever seen from OAI. Serious suggestion: maybe just say something like that on the website. Not buried in a blog post but above the fold.
That is false, considering Astra and Luna don't even exist for 6.1. Terra didn't exist for 6.1 until yesterday. You don't get to make up imaginary tiers. The inconsistency wrt their availability, starting with day 1 of a new model version, is precisely the problem. OpenAI is a user-hostile joke of a firm that users would do well to abandon.
Sure Sol 6.1 on low/medium can be considered "cheap" considering its abilities, but it's clearly not meant for the same type of tasks as Luna. You just grow an intuition for when something is solvable with a Luna-sized model. I'm guessing Terra is good for the tasks that you'd typically use Sol 6.1-low for, or for Luna type tasks (or other not-novel problem) where you want to reduce the risks of suboptimal decision making. It's exactly what I'd go with for feature additions where the correct solution is already mostly designed/specced-out.
Obviously not, otherwise the four models would be called Copilot, as well as the harness.
Luna (Small), Terra (Medium), Sol (Big), Astra (Bigger)
is perfectly clear, sol beats astra sometimes because Astra 6.1 was scrapped. and note that opus 5.5 beats fable 5.1 most of the time too without a skipped model.
"Speed, quality, price: pick two". I think I'd rather see a choice like that than the current convolution of $-per-token, sub quota, latency, simulated reasoning level, model size, context size, etc, etc.
I don't want to rush to their defense here, but, when model families are getting refreshed every 6 weeks, there is going to be significant overlap, particularly with these staggered releases
It is meant to give a rating inside a generation. Different generations are not necessarily comparable. an i7 is better than an i5, but not if you compare an i7 7xxx with an i5 14xxx.
A nvidia xx80 card is better than a xx60 card, but not if you compare a 1080 to a 5060.
The only difference for OpenAI is that not every generation has every model.
I'm optimistic about them tackling frontier intelligence next. Astra 6.1 please?
The "cost per task" metric is just a weaselly way to sound like they didn't just double the cost for most users, and aren't having their proverbial lunch money stolen. They can no longer claim to provide more tokens, more capable models, or more usage in general. So they made up a new metric where "task" could be defined to mean whatever they need it to mean.
They had to become VERY specific in their wording to present any kind of value proposition that will entice people.
[0] https://newsletter.semianalysis.com/p/anthropic-subscription...
Nope to both. Not even close.
The only way it works is if you define "task" in a way that OpenAI is good at it and Anthropic isn't.
Second of all, barring more cuts, yes I do think the $20 and $100 plans do deliver more tasks because the models have gotten cheaper in the meanwhile (like this very announcement).
> Are they delivering 50% more than Anthropic per token?
So say the public benchmarks. Pull up artificial analysis, “Cost (USD) to run all evaluations in the Artificial Analysis Intelligence Index”:
Sol 6.1 (max): $1,082 Opus 5.5 (max): $8,708 Sonnet 5.5 (max): $7,260
>7x difference in cost for the same tasks.
The original plan doesn't exist anymore. To get the same usage you now need a $500/month plan. The new plan that costs $200 is different in every way that counts. It was actually a 150% price increase for people who want or need what the old plan used to provide.
Don't get me wrong, I genuinely LIKE OpenAI. It's probably worth it even at the new price. I just hate the rhetoric designed to Goodhart past the math.
> the models have gotten cheaper in the meanwhile (like this very announcement).
That is not good news for plan users. Plan users now get less of something that also has less worth to begin with. It would only be good news if they got more of it.
For example, if I buy the "unlimited car wash" package for $19.99/month and the car wash lowers the price per wash from $10 to $1 my subscription becomes less valuable, not more. The breakeven has moved, so I now need to wash my car more than 20 times before the package is justified.
If the car wash then increased the price to $50/month would it be wise for me to keep the package? Would lowering the price per wash make the package more valuable to me somehow? No, it would do the opposite.
I’m not sure I see how. Subsidy got cut by half. But in the meantime 6.1 came out which uses 25% fewer tokens than 5.6 and costs 1/3 as much per task.
So for 1/2 the budget you use to have with 5.6 on the old plan, you can do more with 6.1. I don’t see that as a problem.
Once you are inserting models into non-creative (and software development is creative work) it is imperative to start creating your own small evals. Not just to check how different models perform in that equation, but to catch model drift over time.
5x is calculated in terms of API costs, nobody pays those extreme Anthropic prices outside enterprise users.
> like they didn't just double the cost for most users
Most users aren't on their $200 plan. Most users (including on the $200 plan) continue being subsidized.
Ignore value then. Anthropic still provides about 3x as many raw tokens. The 20x\200 plans provide about 10B for openai versus almost 30B for Anthropic.
Do you really think that Opus uses 3x as many tokens as Sol per task? I don't.
> Most users aren't on their $200 plan.
Fair enough. I should have said most heavy users.
> Most users (including on the $200 plan) continue being subsidized.
They continue to be fifty percent less subsidized, I guess.
Look, I don't fault them for trying to become profitable. I fault them for the weird weaselly business speak about cost per task. It's BS game playing and we all know it.
I'm honestly rooting for them, and hope it all works out. For now though I'm pretty salty that my bill jumped by 150% in one month. Gaslighting about made up metrics just makes me more angry.
We agonize over cost per task to the extreme for services/functionality built atop of APIs and can absolutely guarantee you that it is the real world metric that counts most right after “is the model good enough for the task at hand,” and furthermore that, yes, in fact, it is quite often that we see certain models/providers easily using 3x or more tokens on average for the same tasks (or even the exact same task) than a supposedly similarly matched model from another provider. Once you become intimately familiar with the thinking traces from specific models, you can preempt this to some extent with manipulations to the system prompt (essentially heading off entire thinking branches) but the nuking of reasoning traces in the name of preventing against distillation attacks has made this harder.
> Do you really think that Opus uses 3x as many tokens as Sol per task? I don't.
You don’t have to speculate. Pull up artificial analysis, pick either opus or sol and scroll down to total tokens and to total price. Opus (and sonnet) use 4x the tokens and 7x the dollars.
That difference alone removes the difference that semi analysis found.
Fair point. I just tried it, and I stand corrected. Claude uses almost exactly 3x as many tokens.
In the end Claude produced 2.25x more code, and used 3.01x more tokens (ignoring the cached system prompt, etc). It's answer was also much more substantial and feature rich - including many features I did not ask for.
I gave them both the prompt "Without searching the internet please write and test a python script that accepts a year (in any common format) as input and provides the date that easter fell or will fall on in that year. The output should be in discordian date format. You may not use any libraries."
The thing I found most interesting were the unit tests. Claude implemented the core code using both the Meeus–Jones–Butcher algorithm AND Gauss, but only used Gauss as a sanity check to ensure that it's primary code worked. Sol just validated that it matched some dates it could remember. That explains a good chunk of the almost doubled output, and it was a reasonable thing to do, though neither answer was "wrong."
Token usage looked like this (based on parsing it out of the jsonl, neither provider seems to make it easy to get real billing information per project/conversation.
Tool Fresh input Output Cache-excluded Cached reads Gross total
---- ----------- ------ -------------- ------------ -----------
Codex 24,119 7,729 31,848 109,312 141,160
Claude 54,802 41,093 95,895 553,252 649,147I believe oh-my-pi will be able to ingest your Claude and codex conversations, count tokens, and then assign API-equivalent rates to them. May be worth looking into.
GPT-6 Astra
Input:$10.00 / 1M tokens Cached input:$1.00 / 1M tokens Output:$50.00 / 1M tokens
GPT-6.1 Sol
Input:$2.00 / 1M tokens Cached input:$0.10 / 1M tokens Output:$10.00 / 1M tokens
GPT-6.1 Terra
Input:$1.00 / 1M tokens Cache read:$0.05 / 1M tokens Output:$5.00 / 1M tokens
GPT-6 Luna
Input:$0.10 / 1M tokens Cached input:$0.01 / 1M tokens Output:$0.50 / 1M tokens
(For some reason they say "Cache read" instead of "Cached input" for 6.1 Terra)
6.1 Astra and 6.1 Luna must be coming up soon. Will they reduce their prices further?
Interesting that both "none" and "low" reasoning levels appeared not to spend any reasoning tokens at all, and in fact low ended up using less tokens and costing less than none here.
"medium" did use 496 reasoning tokens, but that's apparently not enough here to return a visible reasoning trace.
OpenAI's current frontier model lineup is gpt-6-astra, gpt-6.1-sol, gpt-6.1-terra, gpt-6-luna, so here's a comparison grid of all of their pelicans at different reasoning levels: https://static.simonwillison.net/static/2026/openai-pelicans...
4 pelicans at once!
It's also a very fast model, of all 33 models we've tested so far, only Haiku 5.5 and Nemotron 3.5 Super VL have passed the exam faster but both scored lower. It also wrote the _least_ output in its responses.
[1] Benchmarks for data analytics tasks: https://plotly.com/blog/gpt-6-1-terra-plotly-data-analytics-...
Used on medium seems to be the best use-case[1] and cheaper than 6.1 Sol.
[0]: https://aibenchy.com/compare/openai-gpt-6-1-sol-max/openai-g...
[1]: https://aibenchy.com/compare/openai-gpt-6-1-sol-medium/opena...
The pelicans place it nicely between Sol and Luna in terms of both pricing and results.
The only thing that pelican have is "models where not trained on this" maybe that true months (or years ago) now it's on every HN post and these AI labs are monitoring HN
Luna, Terra, Sol, Astra - are a good setting tweak on how much you want to spend vs what you want to get out of it.
Virtually every task I give to either is better served by going up a tier and down a reasoning level, or down a tier and up a reasoning level. I simply don't understand their use case. No testing I've done has shown them to have a niche where there isn't a cheaper-with-equal-results or more-performant-at-same-cost option.
On subscription for coding and stuff I am basically on Astra Ultra all day long.
2. There are rote or simple tasks where higher reasoning isn't needed and one model vs the other wouldn't even be distinguishable.
3. How many tokens are used during reasoning depends as much on the task as the reasoning level.
4. Sonnet is faster and half the price per token, so Opus will only be cheaper if it can spend fewer tokens. This won't be the case for all tasks. Opus and Sonnet will also consume the same number of tokens to load context, so the relative price comparison depends a lot on where the task is context-heavy or reasoning-heavy.
What are you smoking and can I have some?
I think it's also a fake lever to give to finops / it teams to try and keep costs down.
if you're just talking about use in codex or claude code, then sure, leave it on opus all the time. but for running routine api requests, sonnet is great.
To justify the pricing of Sol and Astra. Old business trick.
If you pay API prices for inference in your product stack & have specific use-cases w/ custom evals, these smaller models provide an additional cost/speed lever to adjust. Ultimately, I think the point is to compete on cost for use-cases where Sol on the lowest reasoning is overkill.
If you have a fairly straightforward usecase where Terra performs just as well as Sol, but is faster and cheaper, why not use it?
Benchmarks don’t really tell you much. For API best will always be a model+effort sweep on your internal evals (and if you don’t have internal evals, do that first).
For individual coding harnesses, honestly I just go with vibes. The best workflow is the one that works for you.
Terra 6.1 on low reasoning outperforms Luna 6 on max reasoning, while costing half as much per task and completing each task over 10x faster (per AA v5).
Luna 6.1 will likely close the intelligence & cost gaps, but that speed advantage will remain.
Spinning up a fresh session would be a reasonable choice in the above case, but sometimes you still need the retained context.
I have some business automation process for some customers that I maintain, where Terra with low reasoning ended up with the best price/performance & latency. With an equivalent Luna with high reasoning, price was usually better, but quality had a higher variance (very much depends on where the thinking takes the model) and latency for high reasoning was worse.
https://docs.github.com/en/copilot/reference/ai-models/suppo...
Take a look at this reply.
GPT-6.1 Terra scored 84, one point over Grok 4.7 (83), in 5.4 min for $0.19 vs Grok’s 36.5 min and $1.69. The cheap models keep getting better.
GPT-6.1 Terra: https://jonclegg.github.io/pacman-bakeoff/#gpt-6.1-terra
Closest competitor (Grok 4.7): https://jonclegg.github.io/pacman-bakeoff/#grok-4.7
I still don't know which gpt model to use! Anthropic one I know as its simple.
Here its gettung weird
Just one thing about the advert, how can I be the only person here thinking the ad is weird - like for the marketing dep, its a bit weird the ad features the same tram that broke and tragically like ended some people - happened for like the first time ever in 100 years, like a year ago ( 14 months ago) just a bit weird.
Im honestly not trying to be negative, but its kinda weird I kinda feel I should mention it because noone else has?
Im sure literally anyone from southern europe must have thought this the second they saw the ad. It was big news last year.
Am I mad? Just a heads up for PR to always check things like this first when releasing a video? Maybe its just me? Just trying to help
When you say its not deep though, I was just saying that literally everyone in Spain a year ago was talking about the yellow tram crash in portugal, and thus it was perhaps insensitive. I dont mind if noone agrees, there was a point to my comment. Just because you live in SF doesnt negate what everyone thought (in Spain, not SF ) when they saw the yellow tram, it was a national tragedy.
I guess weirdly not anymore. man life is crazy sometimes.
I personally like Terra, it kind of does what I would ask Sol or Astra, but dont have to count tokens as much. And Luna has been amazing tbh, it takes designs from prototype and in record time is implementing it, so that was a huge win.
But yeah, they always had confusing way of naming products and if you are not following news, you will be confused about what it is.
is opus 5.5 better than sol 6.1? hmm, i really can't tell. maybe claude is a bit more diligent and identifies more issues up front. but sol has this can do attitude and the codex harness just gets on and does it's job done.
According to this, Sol 6.1 x-high is both cheaper and smarter than Terra 6.1 max.
On AA intelligence index, this fills in the cost per task frontier nicely between 6 Luna and 6.1 sol. So when 6.1 Luna inevitably releases...
The OAI model tango seems to not have an end in sight, for better or worse. Give me more API toys!
(I work at OpenAI.)
My plan for OpenAI recently expired and now I’m thinking to test Anthropic ecosystem. My primary use case is programming only.
Should I go with API use or subscription? For coding harness, I jump between codex, OpenRouter, pi, whatever coding harness is trending on that day
Anthropic does not allow you to use 3rd party harnesses -- you have to use Claude Code. Claude Code is an excellent harness, though.
If you want the freedom to move between harnesses you should use OpenAI or xAI
It's very likely this is a next checkpoint of GPT-6 "Sol", over-optimised for the web chat and offered to Codex subscribers as a "please go away" offering.
If this was half the price, I'd be very interested, but at 1/5/0.05? Feels like another panic release because they can't get Astra 6.1 out of the door...
If not, yawn...
It's nice that their peak-rate hours end quite early in the day UK time.
If you live in a country with a weak currency, DeepSeek models are definitely a godsend.
I still use SOTA for the main model though, but also I'm perfectly aware it's mostly out of FOMO.
Deepseek is more than enough for default subagents though. Any well defined and verifiable task and there's absolutely zero difference between SOTA and good-enough models. Used to be just speed, but Deepseek is fast.
For myself, using Codex, I always have to explicitly tell codex to use Luna with high reasoning for smaller, bit size tasks.
Subagents can spawn with other profiles e.g. scout (uses @smol) and reviewer (@slow). You can set SOTA with higher thinking for @slow and Deepseek with lower thinking for @smol or something even faster/cheaper.
There's also weirdly named /vibe mode that gives explicit instructions to the main model to act as an orchestrator but I find this unnecessary if your main is SOTA, they are already super eager to spawn subagents by default.
e.g.: "You must not use OpenAI models or their outputs to generate, prepare, select, or supply training material or signals for models that are intended as alternatives to or substitutes for OpenAI models or services".
It literally burned through 2 subs.
Unlike Sol, Luna and Astra. Odd.
I said neither of these two statements or claims. The fact is that OpenAI's model family and release is not standardized. This means that if I have applications that are using the high, mid, and low tier models, e.g. at v5.6, it's not straightforward for me to update to a new version, e.g. v6 or v6.1 of these models because there isn't immediately a 1:1 mapping at the time of release of the version. Competitors to OpenAI don't have this problem, so I will bother with them.
Update: Look at the OpenAI employees downvoting me.
Previous Terra model was so good it was discontinued.
OpenAi needs a model to compete with Opus 5.5, and that's not Terra 6.1.
let's see if it can pass the Pangram test. Many services purport to fool it, but fail.
You wouldn't even need to be able to read the information the watermark encodes to recognize it exists in a piece of text
If they couldn't train a reliable detector with that they wouldn't exist.
https://www.pangram.com/blog/third-party-pangram-evals
Note that while that page is from Pangram, the studies on it are not. If they are excluding studies that show Pangram performs poorly in realistic conditions, please share those studies!
At least unless the Terminal Bench 4.0 score is more indicative than the other ones, because that one shows it clearly worse than 6.1 Sol for both performance and cost.
My expectation is Fable 5.5 is next overall improvement, and my bet is OpenAI's "Bel" will be the next competitive one (or, their lack of safety culture will bite them/us in the process).
There's very few stuff that needs Astra/Opus level intelligence.
Luna 6.1 would be amazing.