Grok 4.6 is the flagship, and it holds half the context of Grok 4.3
grok-4.6 carries a 500k token context window[1]. grok-4.3 and the grok-4.20 models, which came before it, carry 1 million[1]. The newer and more capable model holds less[1].
Per million tokens, grok-4.6 is $2 input, $0.50 cached input and $6 output[1]. grok-4.3 is cheaper at $1.25, $0.20 and $2.50[1]. grok-build-0.1 sits at 256k of context and $1, $0.20 and $2[1].
Every model on the list has a second set of rates that starts at 200k tokens, and xAI is explicit about how it applies: a request whose prompt reaches the threshold is billed at the higher rate for all tokens in the request[1]. On grok-4.6 that means $4, $1 and $12, so the whole request doubles[1].
The cached input rate is worth reading next to the others. At $0.50 against a $2 input, grok-4.6 charges a quarter of the input price for a cache read, where most providers charge a tenth[1].