๐ Kimi-K2.7-Code, our latest coding model, is now released and open-sourced!
๐ท Improved coding & agent performance over K2.6: 21.8% on Kimi Code Bench v2, 11.0% on Program Bench, and 31.5% on MLS Bench Lite.
๐ท Reasoning efficiency: Less overthinking, with 30% lower reasoning-token usage compared to K2.6.
๐ท Long-horizon coding: Improved instruction following, higher end-to-end coding task success rates.
โก๏ธ 6x High-Speed Mode coming soon!
๐ Available today via Kimi API and Kimi Code.
๐ Kimi Code: kimi.com/code
๐ API: platform.moonshot.ai
Introducing Claude Fable 5: a Mythos-class model that weโve made safe for general use.
Its capabilities exceed those of any model weโve ever made generally available.
FYI, this is an insta-reject. Temperature 0 decreases model performance significantly, making all results meaningless. It also clearly indicates that the authors have no idea what they are doing.
>be me
>running GLM-5.1 on ollama
>find LPE in windscribe
>70 million users
>full root
>no sudo
>no compiler
>no passwords
>disclose responsibly
>they say "we already patched it"
>check the patch
>it's not patched
>write PoC anyway
>runs clean on release day
>send email
>cofounder replies personally
"not a full exploit"
"we already knew"
"be mindful"
"we have fancier AI than you"
>mfw they have fancier AI but shipped vulnerable code anyway
>mfw the fix is in a version they haven't released
>mfw their own email confirms all who use linux from >their 70 million users are still vulnerable
>mfw no bounty because i went public
>mfw i knew that going in and did it anyway
>mfw "be mindful"
>i am not mindful
>i am ungovernable