When will MiniMax have a Fable level model?
@olive_jy_song, research lead at @MiniMax_AI, joins The Bench to answer that question, plus M3, open source models closing the gap on Vibe Code Bench, and the truth about 996 culture.
Full episode out now!
7:19 – Inside MiniMax's
Moonshot just made history. Kimi K3 is the #2 overall model on the Vals Index, surpassing GPT 5.6 Sol and only slightly behind Fable 5 on the Vals Index.
Inkling supports a 256k context window, has 256k output tokens, image inputs, and tool calling. We used Thinking Machines' recommended evaluation settings: temperature 1, top-p 1, separate reasoning enabled, and reasoning effort set to 0.99.
Nice post discussing calling for a third party standards body! A fun thought experiment here is: suppose this magically happened tomorrow, and you were designated the third party body in charge of testing frontier models. What concrete set of benchmarks or testing framework
I asked mythos 5 max to predict the entire 2026 World Cup. Here are its predictions. It predicts a Spain x Argentina Final with Spain winning. Lets see how it holds up..
We usually set the hills to climb, but sometimes we climb them ourselves 🏔️
Some of the Vals team did the Dipsea Trail this weekend. Couldn't have asked for a better crew!
Want to join the team?
💼 We're hiring for:
- Member of Technical Staff, Research
- Member of Technical
In case you missed it, this was a big week for model releases.
Grok 4.5, GPT-5.6 Sol, and Muse Spark 1.1 all landed within days of each other, and all three cracked the top 10 on the Vals Index. The pace of progress right now is hard to overstate.
Here's our week in review
