Reactions from ranked influencers
5 posts
IDK exactly what this says about distillation. K3 does often specifically say it is claude-opus-4-5-20251101 (while Claude doesn't), but this could just be from misc pretraining data. It doesn't say it's Mythos/Fable, which is a tiny bit of evidence against distilling from these
The full (AI written) write-up for the (slop?) research is here: https://github.com/rgreenblatt/which_claude_is_k3/blob/main/writeups/write_up.md. This research was done fully autonomously by 4.8 so the results may not be trustworthy. I would have used Fable, but the safeguards triggered :/ (Though maybe 4.8 used Fable subagents?) 2/
This is well known, but I do find that Kimi K3 claims to be Claude disproportionately often. I had Opus 4.8 do some (slop?) research about this. I was curious if K3 would ever claim to be Mythos/Fable, but we don't see this. K3 does claim to be 4.5 while claude models don't. 1/
this is what you might call a fair hit
@RyanGreenblatt slop profile (unique phrases) is stronger evidence of distillation than what the model is called and shows kimi k3 distilled from Claude models
@RyanGreenblatt OK, but Fable doesn't know it's Fable