Grok 4.5 Lands, and AUTO Gets Smarter
Anuma has always been about putting more than one model on your side. This week adds a new flagship, makes AUTO smarter about which model to reach for, and sharpens the multi-model tools around them.
A new flagship, and a smarter AUTO
Grok 4.5 joins as xAI's new flagship, featured in the model picker with Grok 4.3 still a tap away. And if you let Anuma use every model, AUTO now blends the two worlds — leaning on open-source models broadly and reaching for closed-source ones on coding and reasoning — so each request lands on the model best suited to it.
A second opinion, one tap away
Under any answer, the Get a new perspective row now returns a true second opinion: a contrasting model's response streams in beside the original, and both stay visible in a two-up so you can weigh them side by side — instead of losing your place in a fresh thread. It's on for everyone, with nothing to switch on.
Council reads the room
Council Mode — Anuma's multi-model arena — now ships a Summary and Analysis rail. Once two or more models finish, a right-hand panel distills their answers into one synthesized paragraph and scores the group: consensus strength, assertion confidence, reasoning quality, and the uncertainty signals worth a closer look. You can also jump straight into Council from any page.
Slack, and a memory that actually searches
Slack joins Gmail, Drive, Calendar, and X in Connected Apps: search messages, browse channels, read threads, and look people up — and post messages, with every post held for your approval first. Separately, the Memory Vault browse page now runs real semantic search — the same on-device engine as chat — so "the hotel I liked in Tokyo" surfaces the right memory even when you don't remember the exact words you saved.
Also this week
Mobile gains real on-device previews for PDF, video, and text attachments plus a broad visual refresh; long conversations on web open and stream faster; and generated apps and slide decks that could stall now finish reliably.