[cebert]

I am very confused about what the difference between Opus 5 and Fable 5 is now. What is the purpose of having two models that are so similar? The main differences I see are cost and marginal capability, according to the Anthropic-provided benchmarks.

[user43928]

Fable 5 is assumed to be a larger model.

It seems plausible to me that RL improvements allowed Anthropic to improve on Opus 4.8, similar to how OpenAI substantially improved upon GPT 5.5 with 5.6 Sol.

Fable 5.1 and GPT-6 are rumored to launch in August, presumably bringing those improvements to the larger models.

[HarHarVeryFunny]

An Anthropic "leak" back in March said that "'Capybara' is a new name for a new tier of model: larger and more intelligent than our Opus models — which were, until now, our most powerful". A second version of the leak had it referring to Claude Mythos rather than Capybara.

I don't know how systematic Anthropic are about their versioning - I'd have guessed that major version number increases (4.x -> 5.x) reflect different base models (different pre-training runs), in which case Opus 5 would be a distilled version of the Fable 5 base model (but without the cyber exploit post-training), rather than being Opus 4.8 with additional post-training, but who knows? I don't believe Anthropic have said anything about this.

[bonoboTP]

I think these versions are largely about marketing and what image they want to project. Bumping the major version indicates/suggests that it's a bigger change in user value. I don't think the technical details matter here for deciding the versioning.

[HarHarVeryFunny]

I think that's part of it too - and I seem to recall someone from one of the labs saying as much about some past model ("it felt more like an 0.5 version increase"). OTOH it would seem odd to me if the "version 5" models weren't related and Opus 5 was Opus 4.8 with some additional post-training rather than coming from the same base model as Fable 5.

[aesthesia]

Yeah, there's probably not a lot systematic behind Anthropic's version numbers. Opus 4.5 was a third of the price of Opus 4.1, indicating there was probably a change in underlying architecture. Opus 4.7 changed tokenizers, probably another base model change.

[HarHarVeryFunny]

Yep, it does seem that way. FWIW I was asking Claude yesterday about the relationships between these different "tier" Anthropic models, and it is not aware of any fixed relationship between them, so it's quite possible that Opus 5 and Fable 5 are related (or not), but that Sonnet 5 is not. Claude suggested that Sonnet 5 was just named Sonnet 5 rather than Sonnet 4.7 since this seemed more appropriate given it's performance relative to Opus 4.8.

OpenAI's versioning seems equally opaque. Claude mentioned that there was a tiny bit of clarity from them in that GPT 5.5 was the result of a new pre-training run, which would seem to strongly suggest that 5.6 (following so soon after, and given the choice of naming) is therefore based on 5.5 - but again who knows.

It seems that so much of the model performance is now coming from post-training that this is what is driving inter-version performance differences, and that base models are much less important than they used to be.

[cesarvarela]

Benchmarks don't reflect the difference between Opus and Fable; you need to talk to them, and eventually you'll be able to tell which one is which without looking.

I think the best proxy for this feeling is the Artificial Analysis' omniscience index. Fable has a 40 score, and Opus (4.8) has 27.

[modeless]

Opus is cheaper than Fable. They could probably replace Fable with Opus but why? They would be churning customers to different models for no reason. Even if a model scores better on benchmarks it can always regress in your specific use case, and customers don't like that. Customers want to be able to continue using their current model until they decide to upgrade themselves.

[Tenoke]

Fable has more parameters. In practice it's not yet clear which one would be better for different usecases yet but they are more different than one being strictly better.

[trentor]

I guess character? Fable is more friendly and curious while opus is a bit more deliberate and conservative.

[dbbk]

Surely that's tunable? OpenAI lets you tune response characteristics.

[trentor]

Yes, but there is a baseline.

[logicchains]

Fable is better for some really hard tasks, the same way it's better than GPT 5.6 Sol, because it's a bigger model.