Third-Party Review: Kimi K3 Agent’s Tasks Are Catching Up to Fable 5, But Hallucinations Are Increasing Too

According to Beating monitoring, Artificial Analysis released an independent review of Kimi K3. It scored 57 points on the Intelligence Index v4.1. Its overall level is comparable to Claude Opus 4.8 and GPT-5.5, but still lower than Claude Fable 5 and GPT-5.6 Sol. This composite index consists of 9 tests, covering real-world work, tool calls, programming, reasoning, knowledge, and long-context.

Kimi K3’s Agent performance is more prominent. In the real-work test GDPval-AA v2, it scored 1668 Elo, surpassing Opus 4.8, GPT-5.5, and GLM-5.2. In the automated SaaS workflow test, the score was 53%, ranking first. In long-form knowledge work, it ranked second, only behind Fable 5.

In this index, Kimi K3’s average cost per task is $0.94. GPT-5.6 Sol is $1.04, and Opus 4.8 is $1.80. However, its hallucination rate in knowledge reliability tests rose from 39% to 51%, while accuracy rose from 33% to 46%.
View Original
This page may contain third-party content, which is provided for information purposes only (not representations/warranties) and should not be considered as an endorsement of its views by Gate, nor as financial or professional advice. See Disclaimer for details.
  • Reward
  • Comment
  • Repost
  • Share
Comment
Add a comment
Add a comment
No comments
  • Pinned