I tested 9 local models on the same flight sim prompt, all Q8, different Q providers, MLX
I gave 9 local models the same flight combat sim prompt. The results broke a few of my assumptions about quant providers and parameter count.
All 8-bit MLX, M3 Max 128GB, served via omlx, prompted through Claude Code. Same prompt every time — single-file HTML, three selectable planes (jet, prop, wildcard of the model's choice), dynamic enemies, tracers, damage, crash spiral on loss. Counted prompts-to-final and graded on "does it actually play."
https://alextzk.github.io/flight-combat-llm-comp
The last time OpenAI did an unannounced livestream, they dropped GPT-4 Turbo and changed pricing overnight
觀點看過了
ChatGPT 畫圖功能大升級!2.0 版本現在直播發表
欸你知道嗎,ChatGPT 的畫圖功能出新版本了!OpenAI 推出 ChatGPT Images 2.0,聽說生成圖片的品質和功能都有大幅進步。如果你有在用 AI 畫圖,或是一直想試試看的話,現在正好可以看直播了解新功能是什麼。簡單來說就是 AI 畫圖又更強了,說真的有點扯。想看現場演示的話,直接上 https://openai.com/live/ 就能看直播。
OpenAI is positioning this as a direct competitor to established image generation tools, suggesting they're confident enough to challenge the current market leaders