xAI released Grok 4.5 on July 8. As of this morning it’s in the Any AI Studio model picker — same dropdown as Claude Opus 4.8, GPT-5.5 and Gemini 3.1 Pro, same credits, no separate signup, no new API key to paste anywhere.

That “one day later” turnaround is the whole point of an aggregator. You didn’t have to notice the launch, read the announcement, create an xAI account, top up a wallet, and wire up billing. You opened the picker and it was there.

What actually changed

Grok 4 already sat at the top of the coding leaderboards for most of the spring. 4.5 widens that lead. In our first day of use the difference shows up in two places:

  • Long, messy refactors. Hand it a 400-line diff with three interacting bugs and it holds all three in its head at once instead of fixing one and quietly reintroducing another. This was the failure mode that used to send us back to Opus.
  • Tool-heavy agent runs. It’s noticeably more decisive about when to call a tool versus reason it out, which means fewer of those loops where the model narrates what it could do for four turns before doing anything.

It’s not the best at everything — that’s not a thing anymore in 2026 (more on that in our model map). Gemini 3.1 Pro still reads a 200-page PDF better, and Claude still writes the most natural prose. But for code and agentic work, Grok 4.5 is the one we reach for first this week.

The point isn’t Grok

Here’s the part we care about. The frontier shipped seven major models between February and April, then Opus 4.8 in May, Fable 5 in June, and Grok 4.5 this week. If your workflow lives inside one provider’s app, every one of those launches is a decision: do I switch? do I re-learn the UI? do I move my history?

Inside the studio it’s none of those. It’s a line in a dropdown. You send the same prompt to the new model, keep the answer if it’s better, and your entire conversation history, your skills, and your uploaded files come along untouched.

Try it: open a thread you already have, hit @ to switch models mid-thread, pick Grok 4.5, and ask it to re-answer your last question. If you like it, keep going. If you don’t, switch back — the previous answer is still right there in the branch.

Credits, not contracts

Grok 4.5 runs on the same credit pool as every other model. There’s no premium tier gate, no “this model costs extra” surprise, no separate subscription. Your Pro plan already covers it, and when the next model lands — and something always lands — it’ll be in the picker too, on the same terms.

That’s the deal: you subscribe to the frontier, not to a vendor.