Kimi K3 lands at Intelligence Index 57 — ahead of Opus 4.8, behind Fable 5

Latent Space (AINews)·high signal

AINews published hard placement numbers for Kimi K3: Intelligence Index 57, versus Claude Fable 5 at 60 and Claude Opus 4.8 at 56, with a Coding Agent Index of 57 that matches GPT-5.6 Terra and GPT-5.5. It ranks #3 among open-weight models on DeepSWE. The framing — 'K3 narrows the gap enough to pressure US labs to ship faster' — is the first quantified read on how close an open-weight model has gotten to the closed frontier.