It is cheaper per token, but it seems to reason a lot more, leading to costs similar to 4.20, but performance is better (similar to what 4.20 had[0]). Overall,

It is cheaper per token, but it seems to reason a lot more, leading to costs similar to 4.20, but performance is better (similar to what 4.20 had[0]). Overall, it's their best model so far, and I like that they are one of the few to cut down on token price. [0]: https://aibenchy.com/compare/x-ai-grok-4-20-medium/x-ai-grok-4-20-beta-medium/x-ai-grok-4-3-medium/ https://aibenchy.com/compare/x-ai-grok-4-20-medium/x-ai-grok...