This is a very particular use case/test, but my first prompt on a new model is always "write a solo fingerstyle guitar tab that blends ragtime, bluegrass, and gypsy jazz".
This is the first model that has responded with something that isn't just a boring arpeggio of chords, so from my perspective it's off to a good start.
Claude Fable 5
161–170 of 1001 posts
Re: Claude Fable 5
#162> A new data retention policy
> Finally, we’re making a change to the way we handle business customer data for Fable 5, Mythos 5, and future models with similar or higher capability levels. We will require 30-day retention for all traffic on Mythos-class models, on both first- and third-party surfaces. We won’t use this data to train new Claude models, or for any non-safety-related purpose, and we’ve instituted new privacy protections including logging all human access to the data and ensuring its deletion after 30 days in almost all cases (see this post for further details). The data will help us defend against complex and novel attacks (including new jailbreaks and attacks that operate across many requests) as well as help us identify and reduce false positives.
Re: Claude Fable 5
#163> We’ve therefore launched the model with safeguards that mean queries on some topics will instead receive a response from our next-most-capable model, Claude Opus 4.8. To release the model both safely and quickly, we’ve tuned these safeguards conservatively—they’ll sometimes catch harmless requests, though they trigger, on average, in less than 5% of sessions. With more capable models arriving in the coming months...
This sounds suspiciously like a capacity story masquerading as a safety story.
Re: Claude Fable 5
#164Is this "system card" equivalent to the stone tablets handed down to Moses? Why don't you call it "user manual"?
Do people chant the "system manual" at Anthropic Tupperware parties? Do they intone a mantra invoking Amodei's name?
Re: Claude Fable 5
#165If you are not seeing it under /model, do a /exit , then a Claude upgrade, then /model again and it should be there.
Re: Claude Fable 5
#166> To release the model both safely and quickly, we’ve tuned these safeguards conservatively—they’ll sometimes catch harmless requests, though they trigger, on average, in less than 5% of sessions.
While I appreciate being conservative, ~5% at the scale Anthropic is operating at is too massive a number. Speaking from my own experience, the actual number is higher than that as well (working on pretty benign tasks such as porting an old open source game into a different language). Opus 4.8 itself even identifies the gaurd's false-positives when its sub-agents are being blocked.
Re: Claude Fable 5
#167At this point Anthropic is a pure marketing and PR company. Super catchy names like Opus, Mythos and Fable trying to get you to think that these software products are actually super-human life changing experiences. Boris Cherny coming to HN “Hi! it’s Boris from the Claude Code team” to get real tech people’s goodwill. From Opus 4.6 there are no noticeable improvements for me in code generation. It works very well, ti…
[dead]
Re: Claude Fable 5
#168> On June 23, we’ll remove Fable 5 from those plans. Using it after that will require usage credits. We've entered the phase where only companies will be able to afford state-of-the-art models.
I hear you, but with the hype surrounding Mythos the demand is going to be insane. I'm already hitting server errors in claude code.
Re: Claude Fable 5
#169I'm very suspicious as they sent out an "We're updating our Privacy Policy" email right before the launch. I fear they try to take advantage of their market position by doing things with user data no other company could do because they know users don't have another choice.
Prob related to this part of the blog post:
> We will require 30-day retention for all traffic on Mythos-class models, on both first- and third-party surfaces. We won’t use this data to train new Claude models, or for any non-safety-related purpose, and we’ve instituted new privacy protections including logging all human access to the data and ensuring its deletion after 30 days in almost all cases (see this post for further details). The data will help us defend against complex and novel attacks (including new jailbreaks and attacks that operate across many requests) as well as help us identify and reduce false positives.
Re: Claude Fable 5
#170I've been running Opus 4.8 for agentic coding and I don't see it being significantly better than Sonnet 4.5 (not that I can tell). I find that pairing Google Gemini and Claude (having Gemini review Claude's code) seems to yield better results. Curious if this jump to 80.3% score in agentic coding will make me see a big difference in actual usage.
SWE-Bench measures single tasks in isolation. In a real loop the model usually loses track of what I was trying to do long before code quality becomes the issue.