What's your mindset around model releases? -- Grok 4.6

Grok 4.6 came in one point behind Fable 5 and level with Sol, at about a third of the price. I

have read the numbers, and I still cannot tell you whether it would change a single thing about

my week.

- The gap is small - one point on the index, and agentic scores close enough to Fable 5 to be

statistically indistinguishable.

- The price is not - two dollars in and six out per million, against five and twenty five for

Opus 5, and roughly half the turns on long tasks.

- The benchmarks disagree with each other - 88 percent on one terminal benchmark and 26 percent

on the next version of it.

Most models now do the same things with minor gains between them, and the variation in what you

get out of one is dominated by how you use it rather than by which one it is. Your orchestration,

your context, the way you write. That is a bigger lever than any two point difference, and it is

the one that is actually yours.

I still keep on top of releases, but for the landscape rather than the decision. I thought about

signing up to Grok to try it and could not find a reason beyond curiosity.

What is your approach when one of these lands?

11 comments

What's your mindset around model releases? -- Grok 4.6

skool.com/cliefnotes

What we give away free beats most paid courses. Build durable AI systems with a Marine vet and Edinburgh researcher. 40+ lessons, growing.

Leaderboard (30-day)

+1133

+844

+657

+567

+514