@TokenGremlin I don't think Astra 6.1 is a good response to Opus 5.5, Opus is good not just because it's smart, but also because it uses a lot less usage. which is the opposite of Astra
CODEX PLUS ACCOUNTS ARE SO BACK BABYYYYY
GPT 6 Sol on High, after reaching full context then 1 compaction and 37K context used.
GPT 6 Luna on Extra High, 122K context used.
and after all this, it only used 22% of my 5 HOUR LIMIT
NOT WEKLY, 5 HOUR.
@needmorevram I hope you beat me to it. I’ll only be able to do it by the end of the week. If it really turns out to be that good, it’ll take the medium-sized-model VRAM crown from Qwen3.8 Flash.
@TokenGremlin Of course nobody is gonna use GPT-6 Luna on Low, but goddamn, that's a new record-low price we've never seen before. Not even DeepSeek ever got this cheap. Insane stuff. Now this is what “too cheap to matter” actually looks like.
Of course nobody is gonna use GPT-6 Luna on Low, but goddamn, that's a new record-low price we've never seen before. Not even DeepSeek ever got this cheap. Insane stuff. Now this is what “too cheap to matter” actually looks like.
bro @zeddotdev did you really to this day never updated your fucking models? Fable 5.1, Astra, the cheper price on Luna and Terra, for fuck sake it has been fucking ages since Luna and Terra got cheaper and your still fucking billing me MORE then their real prices