Moonshot's open-source Kimi K3 model beats Anthropic's Fable 5 on this benchmark
Summary
Our AI Model Release Tracker keeps new models in context with their peers, so you know which are worth your time.
Discussion on
Trending posts from X.
My bet: @thinkymachines will soon make more money than @AnthropicAI. Not by winning the race to build one standardized frontier model. By becoming the Palantir FDE for enterprise custom models.
— Xiaoyin Qu (@quxiaoyin) July 16, 2026
The playbook:
1. Release the best American open-weight model.
2. Drive widespread…
I saw people ask why can't anthropic do this?
— Xiaoyin Qu (@quxiaoyin) July 17, 2026
The short answer is, you cannot both sell closed model API and sell custom enterprise models.
- If you sell closed-model API, you cannot release your weights to enterprises and fine tuning requires Anthropic to run the inference,… https://t.co/HdzTdXoqEb
My initial reaction to Kimi 3:
— Patrick Moorhead (@PatrickMoorhead) July 17, 2026
- it’s an over-reaction shockingly similar the DeepSeek panic
- models like Kimi 3 accelerate and grow the inference market faster than without
- Kimi 3 distilled Fable and therefore big training infra still required for leading edge. No Fable, no… pic.twitter.com/RcxLcDQrss
"Kimi 3 distilled Fable and therefore big training infra still required for leading edge."
— Zephyr (@zephyr_z9) July 17, 2026
Fable was unbanned on July 1
Generating data, cleaning data, and completing the post-training in 15 days isn't possible https://t.co/uvfhBe6qwb