Tibo 在 2026年9月6日的推文

回复 @BeingImmu

RESTLE?

146 回复 · 7 转推 · 2,319 赞

回复 @Arn0Stride7

@Arn0Stride7 Update! Always update!

42 回复 · 2 转推 · 628 赞

回复 @thekitze

@thekitze Why would we

37 回复 · 6 转推 · 853 赞

We've made some improvements that improve usage on the long tail for power users of Astra when logged in with your ChatGPT account. No change in quality and a pure win that on the long tail can result in up to 3-4X less usage being drawn from the subscription.

2,599 回复 · 581 转推 · 16,601 赞

回复 @thsottiaux

The number of people who don't get it in the comments is concerning to me. Go ask Astra.

159 回复 · 12 转推 · 1,608 赞

回复 @Sujay__Raj

@Sujay__Raj @gadi_neelesh @oneill_c I can

12 回复 · 0 转推 · 86 赞

Astra on NVIDIA GPUs every morning: You're not talking to somebody who woke up a loser. That loser attitude, that loser premise makes no sense to me. We are not a car. Brrrrrrr

462 回复 · 186 转推 · 7,530 赞

回复 @gadi_neelesh

@gadi_neelesh @oneill_c I have a rule for Twitter. I only hand-type things. Neurons to letters, with only the keyboard inbetween

13 回复 · 2 转推 · 165 赞

回复 @Gelassoldat

Who says it won't reset in a while 👀

301 回复 · 63 转推 · 1,580 赞

回复 @oneill_c

Researcher. They spend their time in the most incomprehensible pile of Python code that requires having read 10 different papers and the full comprehension of the very odd systems and how they subtly affect training dynamics. Their entire day is basically debugging why something doesn't work, not knowing if it's the math that's wrong or if the system is wrong. Among engineering and product, the picture is very different and we started to see massive uptick months before.

48 回复 · 38 转推 · 1,875 赞

回复 @pfernan95dev

@pfernan95dev Welcome back Pablo. See DM for a little gift.

164 回复 · 6 转推 · 1,762 赞

回复 @TJeparskis

@TJeparskis Should not be the case, between those two it's about ~50% cheaper per task at roughly equivalent output.

473 回复 · 14 转推 · 1,784 赞

To calibrate you all on which reasoning effort to use for Astra, know that GPT-6 Astra on low performs better than GPT-5.6 Sol on high. If you were using high reasoning efforts with Sol and were happy, I suggest you move down to low or medium for Astra.

1,834 回复 · 1,076 转推 · 22,007 赞

回复 @hiarun02

Not the case. There is no difference between usage you get before or after a reset.

194 回复 · 31 转推 · 2,313 赞

回复 @FelipeFr1702

@FelipeFr1702 You’re welcome

164 回复 · 35 转推 · 4,843 赞