YouTubeFeed

Claude Fable 5 - is this Mythos model worth the wait?

Summary

Claire Vo got early access to Claude Fable 5 — the first Mythos-class intelligence model to reach general availability — and puts it through real work before giving her verdict. Anthropic frames Fable as a completely new model class above Sonnet and Opus: state-of-the-art on SWE-bench Pro, capable of days-long autonomous sessions, and priced accordingly at $10/$50 per million input/output tokens. It’s also gated by new safety classifiers for cybersecurity, biology, chemistry, and distillation that gracefully fall back to Opus 4.8 rather than refusing outright, and Vo confirms this is technically Fable, not the more restricted “big Mythos” still limited to Anthropic’s Project Glass Wing partners.

Her hands-on verdict is a mixed bag. Fable genuinely impressed her on vision tasks — PDF and document formatting came out cleaner and more readable than Opus 3 — but its prose is “nearly unreadable” for specs and PRDs, over-engineered and hard to parse the way an overly thorough senior engineer’s writeup can be. Front-end design was a bigger shock: one-shot UI output was flatly bad, even after more detailed prompting, which Anthropic support suggested was needed. She also ran into real friction testing Fable’s promised long-running, multi-agent orchestration, including a batch of seven sub-agents that silently stalled for three hours while she was away from her laptop.

Her closing recommendation is to reserve Fable for genuinely hard, detail-critical technical problems and vision work, while routing spec writing, prose, and front-end design to Sonnet or Opus. She doesn’t come away a hater — she just thinks the model needs a narrower, more deliberate slot in the stack rather than being reached for by default.

Highlights

”The model, the myth, the legend”

Mythos has arrived

“It’s here. The model, the myth, the legend. Mythos from Anthropic has finally dropped. Well, baby Mythos. We’re calling it Fable 5.” — Claire Vo, 0:00

Clip command
yt-dlp --download-sections "*0:00-0:31" "https://www.youtube.com/watch?v=IREnr4I89Ho" --force-keyframes-at-cuts --merge-output-format mp4 -o "model-myth-legend.mp4"

Graceful fallback: safety classifiers punt to Opus 4.8, not a refusal

Fallback instead of refusal

“If you get classified into one of these categories, instead of saying like, do not pass go, you may no longer [use] stable, it just falls you back to Opus 4.8.” — Claire Vo, 6:27

Clip command
yt-dlp --download-sections "*6:27-7:21" "https://www.youtube.com/watch?v=IREnr4I89Ho" --force-keyframes-at-cuts --merge-output-format mp4 -o "graceful-fallback.mp4"

”The writing is nearly unreadable”

Nearly unreadable prose

“Here’s the problem though. The writing is nearly unreadable. So if you’re thinking about Claude for prose, for spec writing, for PRDs, unfortunately it’s an engineer and what’s the problem with engineers? They just really get wrapped around the axle on details.” — Claire Vo, 11:14

Clip command
yt-dlp --download-sections "*11:14-12:00" "https://www.youtube.com/watch?v=IREnr4I89Ho" --force-keyframes-at-cuts --merge-output-format mp4 -o "nearly-unreadable.mp4"

”Legitimately terribly bad” at one-shot design

Design shock

“The other thing that shock, shock, shocked me was how like, actually legitimately terribly bad it was at design or at least at one-shot design. I mean, I’m not even talking like AI slop bad, it’s like fundamentally terrible design: gray, black, red, simple outlines, just really, really terrible.” — Claire Vo, 12:56

Clip command
yt-dlp --download-sections "*12:56-13:51" "https://www.youtube.com/watch?v=IREnr4I89Ho" --force-keyframes-at-cuts --merge-output-format mp4 -o "terrible-design.mp4"

Seven sub-agents, silently stalled for three hours

Seven stalled agents

“Now, I made the mistake. I walked away from my laptop and came back to these seven agents that had stalled after about three hours. And so like, egg on my face, but I really want to see how technically the Claude code model holds up to the promise of multi-agent orchestration.” — Claire Vo, 15:08

Clip command
yt-dlp --download-sections "*15:08-15:41" "https://www.youtube.com/watch?v=IREnr4I89Ho" --force-keyframes-at-cuts --merge-output-format mp4 -o "seven-stalled-agents.mp4"

The verdict: hand it hard problems, not your front end

Where Fable fits

“I probably wouldn’t hand it my front-end work, or I definitely wouldn’t hand it my front-end work, and I definitely wouldn’t hand it strategy or spec work. I think it overthinks things. I think its prose is nearly imparsable.” — Claire Vo, 15:41

Clip command
yt-dlp --download-sections "*15:41-16:25" "https://www.youtube.com/watch?v=IREnr4I89Ho" --force-keyframes-at-cuts --merge-output-format mp4 -o "verdict-hard-problems.mp4"

Key Points

  • Fable 5 is “baby Mythos” (0:31) - It’s the first Mythos-class model at general availability; full “big Mythos” is still restricted to Anthropic’s Project Glass Wing partners
  • Pricing: $10/$50 per million tokens (0:31) - A new tier above Opus; not cheap
  • A completely new model class (1:54) - Sonnet, Opus, and now Mythos; Anthropic claims state-of-the-art benchmark results, including ~80% on SWE-bench Pro
  • Built for long, autonomous tasks (2:22) - Anthropic markets it as capable of days-long asynchronous work, proactive behavior, and exceptional vision
  • Token consumption is ~2x (3:00) - Anthropic says Fable consumes rate limits and tokens at roughly double the rate of other models
  • “Works like a seasoned engineer” — pros and cons (4:15) - Extremely thorough and complete, but that thoroughness isn’t always in service of shipping or building a great product
  • Extra-high effort mode used throughout (5:11) - Vo deliberately tested on the highest reasoning/effort setting to give the model its best shot, despite Anthropic recommending “high” for most work
  • Safety classifiers for cybersecurity, biology, chemistry, distillation (6:27) - Flagged sessions gracefully fall back to Opus 4.8 rather than refusing; 95% of sessions never hit a fallback
  • 30-day retention policy (7:21) - Used only to catch misuse, not for training Claude
  • Is this Mythos? Yes, technically (7:49) - Fable and Mythos are the same underlying model; Fable simply ships with the public safeguards attached
  • New product launches alongside Fable (8:34) - Claude Managed Agents enters public beta; a new “senior advisor” strategy pairs Fable with cheaper execution-layer models
  • Vision is a standout strength (10:14) - Document and PDF formatting (tested on handwriting worksheets for her kid) was cleaner and better laid out than Opus 3
  • Prose is nearly unreadable (11:14) - Adversarial review of a ChatPRD product-graph spec came back long, detailed, and internally consistent but very hard to parse
  • One-shot design was a real disappointment (12:56) - A skills-registry UI came out flatly bad even after Anthropic’s suggestion to prompt with more detail
  • Overly conservative on “MVP” scope (14:08) - Asked to ship a customer-value MVP, Fable delivered something too narrow and not very useful
  • Multi-agent orchestration: real but shaky (14:57) - Some successful sub-agent runs, but also stalls and errors, including seven agents that silently stalled for three hours
  • Final verdict: narrow, deliberate use cases (15:41) - Hand it hard technical and vision problems; keep spec writing and front-end design on Sonnet or Opus
  • Not a hater, still finding its place (16:21) - Vo plans to keep testing Fable rather than writing it off

Mentions

Companies

  • Anthropic (0:31) - Maker of Claude Fable 5 / Mythos; source of all benchmark and safety claims discussed

Products & Technologies

  • Claude Fable 5 / Mythos (0:00) - The new model under review; first Mythos-class model at general availability
  • SWE-bench Pro (1:54) - Benchmark where Anthropic claims ~80% state-of-the-art performance
  • Opus 4.8 (6:27) - The model safety-classified sessions gracefully fall back to
  • Claude Managed Agents (8:34) - Anthropic’s hosted harness/sandbox for long-running agentic work, entering public beta alongside Fable
  • Claude Code (14:57) - Used to test Fable’s multi-agent, dynamic-workflow orchestration
  • ChatPRD (11:14) - Vo’s product; its open-source “product graph” spec was the test subject for Fable’s writing/review capability

People

  • Claire Vo (0:00) - Host of How I AI, founder of ChatPRD; solo review and hands-on tester in this episode

Surprising Quotes

“Your girl’s working on PRDs, she’s shipping SaaS, she’s not working on biology quite yet, although give me a little time and some time to experiment maybe I’ll get there.” — Claire Vo, 0:31

“Sometimes you want it to be a little bit dumber.” — Claire Vo, 4:15

“It’s just really hard to see the forest for the trees in this particular model.” — Claire Vo, 12:21

“I cannot wait to hear what you build, what you overbuild, and what you make ugly with this new model.” — Claire Vo, 16:43

Transcript

Claire Vo: 0:00 It’s here. The model, the myth, the legend. Mythos from Anthropic has finally dropped. Well, baby Mythos. We’re calling it Fable 5 and this new model is crushing benchmarks but the question is, can it crush my backlog? I got early access to the model and of course I have my own opinions on where it does really well, where it needs a little work, and the question on everyone’s mind, does it live up to the terrifying marketing hype? Let’s get to it.

Claire Vo: 0:31 Okay, let’s talk about what Anthropic is telling us about this model and then we’ll get into what I think about it. So this is Claude Fable 5, the first Mythos class intelligence model to reach GA. Now, if you haven’t been paying attention, Anthropic has been marketing/scaring/warning us about the unbelievable capabilities of Mythos and it is finally here. Now they had originally been rolling this out with a couple select companies. I got early access to test drive what I thought of the model, but you have to know, this is not Mythos capital M, big Mythos. This is baby Mythos. This is Fable and so it’s going to have some guardrails on it, in particular around cybersecurity exercises and biology exercises. Now, good news, your girl’s working on PRDs, she’s shipping SaaS, she’s not working on biology quite yet, although give me give me a little time and some time to experiment maybe I’ll get there. So this is really going to be focused on what the everyday user with the everyday software engineer is going to think about when they’re using this model. Although I did run into some things that I suspect are a result of the tuning and training of this particular model to be extra safe. Now, quick, it’s not cheap. It’s $10 per input token and $50 per output token. It’s going to be a new tier above Opus and so if you’re going to use this model, you’re going to pay pay the price.

Claire Vo: 1:54 So what is Anthropic saying? Basically it’s a completely new model class. So we had Sonnet, we had Opus and now we have Mythos, the first of which is Fable 5. It’s completely state-of-the-art. It is exceeding every benchmark they tested by a significant amount. This 80% on SWE-bench Pro, you’ll look at that compared to some of the more recent models that have come out. Very, very good benchmark performance. And then they’re saying it’s really good for long complex tasks.

Claire Vo: 2:22 Now, what are some things that earlier models couldn’t do that they are saying now that Fable 5 can do? It’s very autonomous, including running days-long asynchronous tasks. It’s really an engineer’s engineer and at some of the downside I experienced with this model I’m going to show you a very specific example of where you don’t want an engineer doing your work with an engineer’s point of view. Proactive, um, it’s very good at vision, exceptionally good at vision. This is a place where I actually really loved the model and you know me I’m pretty critical of models but I did see a step ahead of vision so that’s something we’re going to dive into and then effort. It works hard, it builds harder, it verifies more. It’s really a different level of autonomous coding and we’re going to see how that plays out in some of my research.

Claire Vo: 3:00 Thought for ambitious work. Now, guess what it also can do? It can consume those tokens. So, Anthropic has said it consumes rate limits and tokens at about 2X the rate of other models, so again, this is a big boy model and it’s going to consume tokens and some of the things that it’s good at and even some things that they have done in the harness seem like they are intentionally or not token consumers. So, we’re going to keep an eye out on cost and an eye out on efficiency when using this model.

Claire Vo: 3:30 Again, talking about long-running tasks, Fable 5 is supposed to be able to run for days. So, doing long-running planning, being able to spin up sub-agents and I show a little bit about dynamic workflows, which are, you know, different architectures of sub-agents and holding multi-day sessions. Now, I have done probably day, days long sessions with other models. I didn’t have Fable for many days, so I cannot verify that it ran for days. I did get it to run, however, for several hours on some tasks that may or may not have merited that several hour effort, but it definitely seems like it has both the harness and the intelligence capability to run for a very long time, if that’s appropriate for your task.

Claire Vo: 4:15 Now here’s your pros and here’s your cons. They explicitly say that Fable works like a seasoned engineer. Unfortunately, if you have worked with a seasoned engineer, you know there’s good to this and you know there’s bad to this. So, it is very complete in its investigation and it’s definitely going to go search out all the corners. It’s definitely going to think about how it can be 120% sure that it’s shipping the right thing. But guess what? That’s not always in service of launching and that’s honestly not always in service of building a great product. So while you can give it a goal and it will be very autonomous and it will be very thorough, honestly sometimes you want like a slightly less thorough engineer. Product manager talking, even engineer talking. Sometimes you want it to be a little bit dumber. We’ll talk about some of the prompting techniques it says and when to use this model, but it’s just something to think about when you’re working with any high intelligence model is how much intelligence does the task actually take?

Claire Vo: 5:11 Now, as I said before, it is token intensive by design and I did most of my tasks on extra high and so it was like token burning on token burning. And so they say that high is probably the sweet spot for most work. I used extra high just because I don’t want anybody in the comments saying, ‘Claire, you picked high for this task and it should have been extra high and you would have had a better experience.’ I used extra high. I used all of the brains of Fable. But again, it is very, very token intensive. My question for any of these models, this is not an Anthropic model question, this is not a Fable question, is does this token intensity actually output the right result? And that’s a place where I’ve just not 100% sure. But again, as a…

Claire Vo: 6:00 As humans in the loop, we’re going to have to be much more intelligent about where to put what model and where to use what reasoning and what effort level to match what we’re doing. And again, I think that the untrained of us will say, oh, well, I have this stable model, I should use it, it’s better than anything. And honestly, I still think there’s a place for good ol’ Sonnet, I think there’s a place for Opus, and I think there’s a place for other models in the ecosystem.

Claire Vo: 6:27 Now, there are safeguards in this model. And so this was one of the first things that Anthropic told me testing the model, and this is one of the headlines that they’re making in the release, which is there are specific classifiers in this model for cybersecurity, biology, chemistry, and distillation. Basically, they don’t want anybody doing bad stuff in those categories in particular with this very intelligent model. What’s nice about how they’ve implemented this, however, is they have this new fallback concept. And so if you get classified into one of these categories, instead of saying like, do not pass go, you may no longer stable, it just falls you back to Opus 4.8. This is also a capability in the API now where you can do this graceful fallback to 4.8 if you’re using a Mythos class model.

Claire Vo: 7:21 They also have a 30-day retention policy, used only to catch misuse and it’s not used to train Claude. So while it’s still not cleaning training Claude, they do want to check the use of this model because they have been and will forever be very cautious about us normies using their intelligent models. And just, you know, for context, 95% of sessions on this model did not hit a fallback. I don’t believe I hit a fallback, but again, I’m not doing anything in cybersecurity, biology, or chemistry, at least yet.

Claire Vo: 7:49 Okay, so this is the question: is this or is this not Mythos? It is Mythos. Fable has the safeguards, Mythos does not. Fable, all us normies can have in general availability. Mythos is still restricted to these project glass wing partners, some of these enterprise-level partners that are really checking it against cybersecurity use cases. I would suspect that at some point we get some access to a Fable 5.whatever or that the project glass wing class opens up. But for now, we get Fable, project glass wing or these pre-selected companies get Mythos, but they are all fundamentally the same underlying model.

Claire Vo: 8:34 A couple product things that are also launching today along with the Fable 5 model: Claude Manage Agents are going into public beta. If you haven’t paid attention, this is Anthropic’s hosted harness, hosted sandbox for running long-running agentic work. I am still trying to figure out what a good use case for Claude Manage Agents is. I will get there, um, but Fable ships out of the box in Claude Manage Agents. There’s also a new advisor strategy where you can use Fable 5 as a senior advisor.

Claire Vo: 9:00 and use cheaper models as an execution layer. A lot of people are doing this with Opus and Sonnet and so this is going to work today in the API and in Claude code and is a strategy you can use. And then as I mentioned, this fallback API where you can put an optional parameter on the messages API that allows you to continue block requests by using 3.5 at Opus pricing.

Claire Vo: 9:22 Okay, as we said, crushing benchmarks. Look at this, Claude 3.5 compared to Opus 3, GPT-4o and Gemini 1.5 Pro. Significant increase in SweetBench Pro benchmark, um, very far ahead of these other models. And while I wasn’t testing the most advanced use cases, I didn’t find something that technically it failed at. So I think these benchmarks are really going to hold. And these benchmarks have outperformed across the board. So this is Anthropic’s state of the art model.

Claire Vo: 9:57 Okay, so enough about what they say. Let’s talk about what I say. What is it actually like to use? So I ran Claude 3.5 on a bunch of different work and I want to give you my feedback of where I thought it did well, where it needed a little bit of work and where I was really surprised.

Claire Vo: 10:14 As I said before, it’s really good at vision. And where is it good at vision that really impressed me? It’s really good at document formatting. So this is super simple, um, but we’ve been doing these handwriting documents for my seven year old based on classic texts and classic poems. And on the right is Opus 3 and on the left is Claude 3.5 and it looks so silly, but I really do think Claude 3.5 did a much better job of a second grade layout for a handwriting sheet. There’s just like the right spacing, it’s very clear to read. There’s enough white space. I think on the one on the right, it’s just very dense and even the lines themselves are sort of hard to tell, do you write above, do you write below? So I do think that PDF formatting documents I tested this against a bunch of different models, Claude 3.5 really did a good job. So very simple eval for me, but a very, very good one.

Claire Vo: 11:14 Now here’s the problem though. The writing is nearly unreadable. So if you’re thinking about Claude for prose, for spec writing, for PRDs, unfortunately it’s an engineer and what’s the problem with engineers? They just really get wrapped around the axle on details and this is a real struggle with these more intelligent frontier models is they’re too smart. And so it’s just very, very hard to parse what they’re saying. And I’m going to show an example of this in actually Claude code. So I have this concept of a product graph that I’m working on for Chat PRD. It’s actually a fairly complex open source project. And I had Claude 3.5 go through that and actually do like an adversarial review of my requirements to try to figure out…

Claire Vo: 12:00 …out, where there were internal consistencies in the logic and, it gave me this markdown document that looked very long and intelligent. But, when you actually go through it, it’s just really hard to parse. It’s just like internal references, it’s very detailed but not in a way where you can zoom out.

Claire Vo: 12:21 There are these big blocks of paragraphs, like look at, look at this. It is just really hard to see the forest for the trees in this particular model. And I saw this sort of like over and over again working on it with specs is, it was very complete but nearly unparseable.

Claire Vo: 12:41 And that’s a real challenge when working with these very, very high intelligence models. Again, I would actually suggest pulling back to maybe a Sonnet or Opus model for specs and then looking at Fable as an orchestrator of execution where that detail really matters, but you don’t have to read it.

Claire Vo: 12:56 The other thing that shock, shock, shocked me was how like, actually legitimately terribly bad it was at design or at least at one-shot design. And so I asked Fable to design a skills registry and man alive, did it do a very poor job.

Claire Vo: 13:14 I mean, I’m not even talking like AI slop bad, it’s like fundamentally terrible design: gray, black, red, simple outlines, just really, really terrible. Now, the Anthropic team suggested that I just needed to be a little bit more detailed in my prompting.

Claire Vo: 13:31 I’ve never had to do this before in, I would say, the last year of models in terms of front end, but even when I prompted it, it was still just not very impressive design. I think there’s this real balance between design slop and specificity in just shipping like terrible design.

Claire Vo: 13:51 I’m not sure what about Fable 5 resulted in this. I’m going to have to keep testing it as it rolls out today, but this was a real disappointment in terms of design. So again, you might want to toss an Opus in the mix instead of relying on Fable for design.

Claire Vo: 14:08 It’s really conservative on execution. So when I was trying to do that ambitious days-long work, I took a spec and I said, ‘Can you ship the V0 of this, the MVP?’ I said, ‘Enough so that a customer could get value.’ And the MVP, they just really took minimal to heart.

Claire Vo: 14:26 It was like very, very narrow, not actually that useful. And I’m curious if this comes from some of the safeguards on this model and, it’s been a challenge I’ve seen since the kind of later Opus models is they’re not super ambitious.

Claire Vo: 14:41 And so again, you’ll have to think about how to prompt this to get that long-running outcome paired with the right product ambition. And then I really doubled down trying to test these Claude dynamic workflows and these sub-agent designs, trying to see if this would really add value.

Claire Vo: 14:57 And the multi-agent capability is definitely there, and I definitely had some successful multi-

Claire Vo: 15:00 The agent runs kits off in Fable, but I also ran into a lot of stalls and errors in using multi-agent orchestration.

Claire Vo: 15:08 Now, I made the mistake. I walked away from my laptop and came back to these seven agents that had stalled after about three hours.

Claire Vo: 15:16 And so like, egg on my face, but I really want to see how technically the Claude code model holds up to the promise of multi-agent orchestration.

Claire Vo: 15:26 I had some successes and some bugs. I think this is a Claude code issue, not necessarily a model issue.

Claire Vo: 15:32 Although with this promise of long running, days-long prompts, you really got to deliver technically on the outcome.

Claire Vo: 15:41 So, what’s my takeaway? I would hand it hard problems, of course not cybersecurity, bio or chemistry problems, but hard technical problems where being extremely detailed matters.

Claire Vo: 15:49 Long horizon work. I would also hand it vision problems where you really want something to look good or you want it to parse PDFs or other documents. It’s done exceptionally well there. I was actually really surprised.

Claire Vo: 16:00 I probably wouldn’t hand it my front-end work, or I definitely wouldn’t hand it my front-end work, and I definitely wouldn’t hand it strategy or spec work.

Claire Vo: 16:07 I think it overthinks things. I think its prose is nearly imparsable.

Claire Vo: 16:13 And so maybe I’ll test it again with effort level lower on sort of prose and spec writing, but it wasn’t it for that.

Claire Vo: 16:21 That being said, I’m not a hater on this model.

Claire Vo: 16:25 It definitely has a place in your stack. I’m going to test it. If you want to learn more, definitely look up the prompting guide for Fable.

Claire Vo: 16:31 It’s gonna probably repeat a lot of what I said: hand it your hardest problems, what this model’s good for and what it’s not, and how to get a good outcome.

Claire Vo: 16:43 That being said, Mythos is here. I cannot wait to hear what you build, what you overbuild, and what you make ugly with this new model. Thanks for joining How I AI.

Claire Vo: 16:54 Thanks so much for watching. If you enjoyed the show, please like and subscribe here on YouTube, or even better leave us a comment with your thoughts.

Claire Vo: 17:04 You can also find this podcast on Apple Podcasts, Spotify, or your favorite podcast platform.

Claire Vo: 17:12 Please consider leaving us a rating and review, which will help others find the show.

Claire Vo: 17:18 You can see all our episodes and learn more about the show at howiaipod.com. See next time.