dev With the power of AS7018 (AT&T Enterprises, LLC), no further model degradation has occurred.

Claude Pro ran out of quota ➡️ Subscribed to ChatGPT Plus ➡️ GPT-6.1 Sol was unusable / GPT-6 Astra was too expensive ➡️ Had less than 30% of my five-hour quota left ➡️ Resubscribed to Claude Pro ➡️ Still had to wait until Monday at 1:00 p.m. ➡️ Finished the routine using less than 18% of my five-hour quota with GPT-5.6 Sol Low (much smoother than GPT-6.1 Sol High, with only a third of the usage)

Intelligence: GPT-5.6 Sol Low > GPT-6.1 Sol High Cost: GPT-5.6 Sol Low < GPT-6.1 Sol High

Intelligence: GPT-6 Astra Medium >> Opus 5.5 High >> GPT-6 Astra Low Cost: GPT-6 Astra Medium > GPT-6 Astra Low >> Opus 5.5 High

🤢

GPT-5.4 = GPT-5.5 = GPT-5.6 Terra = GPT-6 Sol = GPT-6.1 Sol

dev I will never use GPT-6.1 Sol again for the rest of my life.

If Anthropic was not A/B testing their new shiny quota discrimination technologies, I would never consider OpenAI as an alternative. 🤢

🤢

Barely usable.

dev GPT-6 Astra Low 🤢

Thariq & Boris Cherny >> Tibo

dev Claude Code >> Codex

dev GPT-6 Astra | 7 prompts | 9:46 PM - 9:54 PM CST 12% of the 5-hour limit

アホすぎてイライラ中

そもそも Fable 5.1 や GPT-6 Astra とか高すぎて使う勇気がないよ

Opus 5.5 >> GPT-6.1 Sol

GPT-6.1 Sol is even worse. 🤢