benchr
Two changes, one notice
It is worth separating the friendly half of this from the half that costs money.
The good change: xAI consolidated a confusing menu. Grok 4.3 is now the default for chat, and the legacy aliases — Grok 4, Grok 4 Fast, Grok 4.1 Fast, Grok 3, and Grok 3 Mini — all route to it. You no longer have to reason about which old variant to call; you can just call Grok 4.3 directly. For a hands-on take on the model itself, see our Grok 4.3 review.
The change to check: "redirect" includes billing. Requests to deprecated text slugs are billed at Grok 4.3 pricing, not at whatever the old slug used to cost. If your application was quietly running on a cheaper Fast tier, its per-token cost moved up the moment the redirect took effect.
The pricing, plainly
| Grok 4.3 | Input / 1M | Output / 1M | Cached input / 1M |
|---|---|---|---|
| Standard (≤ 200K input) | $1.25 | $2.50 | $0.20 |
| Over 200K input | 2× standard | 2× standard | — |
By itself, $1.25 / $2.50 is reasonable for a frontier-ish model — it is cheaper on output than most flagships. The problem is relative, not absolute. Apps migrating from Grok 4.1 Fast face roughly a 6.25× jump on input and a 5× jump on output versus the old Fast pricing, simply because the cheaper tier no longer exists as its own billed product. Nothing about your code broke; the line item did.
How to audit your app in ten minutes
- Grep your codebase for hardcoded model strings. Search for
grok-3,grok-4,grok-4-fast,grok-4.1-fast, and similar. Each match is a call that now lands on Grok 4.3. - Check config and environment variables too. Model names are often set outside source — in
.envfiles, deployment configs, or a settings table. - Compare last month's bill to this month's. A step change with no traffic change is the tell that a redirect repriced you.
- Estimate the new cost deliberately. Run your real input/output mix through the cost calculator at Grok 4.3 rates before deciding whether to stay, optimize, or switch.
- Watch the 200K line. If any of your prompts run long, the over-200K 2× multiplier compounds the change.
Should you stay on Grok 4.3?
That depends on what you used Grok for. Its distinguishing feature is live access to the web and to X, which is genuinely useful for "what is happening right now" questions and weak for everything else. If that real-time edge is why you chose it, the new price may still be worth it. If you were on a Fast tier purely for cheap throughput, this is a good moment to compare alternatives — both against other flagships in the AI model pricing comparison and head-to-head in Grok 4.3 vs ChatGPT. You can also line it up against any model side by side in the compare tool.
One honest note on framing: xAI is not raising the price of a product you were buying — it is retiring the cheaper products and pointing their names at a pricier one. The distinction matters for how you respond. There is no "old plan" to negotiate back to; the right move is to re-evaluate fit at the new rate, not to wait for a reversal.