OpenAI has announced GPT-6 Astra, hailed as a generational leap in capability across cybersecurity, science, and software engineering, entering the AGI era.
1 editorial report · 3 verified social mentions. The most authoritative report leads while later evidence completes the story.
TL:DR ... and they should feel so #%&!!! now that even OpenAI is taking actions to restore (up to certain point) the limits to be somehow higher again. ----------- This kind of weird people that acts like shooting at their feets use arguments like: "They was subsidizing the subscriptions for a long time so this is only an adjustment to the reallity". Well, the inference cost is not the problem, the problen is the crazy IA race and the amount of money they spend to get the next best model. But this works this way in many companies investing in I+D: they know the profits will come at long horizon and they invest to be there in a good position when the time comes. "If you are a Plus user you don't have reasons to complain, because no serious developer has a Plus subscription, the seriuos ones
Open mentionIt's Meta's answer to the "AI should be for everyone" debate — and a not-so-subtle jab at closed labs like OpenAI and Anthropic who keep their best models locked away. Zuckerberg's betting big: this is just the appetizer. The full-power Muse Spark model is set to go open-weight next. # metaads # museglimmer # opensourceai # artificialintelligence # ｍodels # technews # zuckerberg # machinelearning # agenticai # openweightai
Open mentionYour best model is the industry's best (at least till we get to see what OpenAI's Astra is like) but it burns tokens like crazy, and on top of that, you cannot offer it full scale due to compute shortages. Your next best model is supposed to be *the* historical workhorse, Opus 5, but is shit. Your best affordable model is historically supposed to be effective, Sonnet 5, but is shit. Your effective next best models are stale flagships, Opus 4.8-4.6 and Sonnet 4.6. I mean, currently; Anthropic offers one true flagship at a nearly unusable scale Anthropic offers a couple other usable models with stale performance Until 6-8 months ago Opus was revered as fuck, it was *the* AI to problem-solve with, it was insightful, it was a workhorse, it was your go-to. It was what forced OpenAI get its shit
Open mention