Zai should have given ox alpha a major model name. GLM 6 (considering they trained a new model), or a variant name like glm 5.3 mythic (similar to sol/terra/luna). As ox alpha they convinced the world they were frontier fable level, then revealed it as a sub variant of an already released model. They will assume lack of usage is due to it previously being free, but rather it will be because people perceive it to be less intelligent. I was convinced ox alpha was smarter than both fable and sol 5.6. Now that I know it’s a flash variant of an existing model that benchmarked below frontier I won’t use it.
Today ox alpha is responding more intelligently than fable.
It’s not being as much of a yes man, and it’s finding nuanced points that fable misses.
Truly an impressive model.
What if ox alpha is sub Q? They wouldn’t need the massive compute for a free week, since they can radically reduce compute required to serve the inference.
If so it would be a glorious comeback for them.
Grok 4.5 is insanely fast and almost as good as opus 4.8 and gpt 5.5. Still a step below fable and sol, but makes up for it with its speed.
Becoming my go to agent in CLI. But its weird that in minimal mode I can't stop a request with esc, or send a queued message immediately.
Meter is an incredibly powerful AI app.
Access Claude, GPT, Grok and Gemini in one place, get them to debate each other on your ideas so you get clarity of thought, and pay per thought, not per month.
Here’s everything it can do.
🧵👇
I don’t dislike the disco ball aesthetic of Spotify’s new logo, it’s the white ring that kills the whole thing. Looks like an app update downloading. How can a public company of this size not flag this in design review. 👎🤯🪩🕺