Grok 4.7 is live, and the fast release rhythm continues. xAI shipped the new model on September 21, 2026. That's roughly a month after Grok 4.6. It arrives at the same price and speed as its predecessor. That combination, more capability for the same money, is the pitch. So is it worth switching? That comes down to what you actually use a model for.
Grok 4.7 Release Details
The Grok 4.7 update lands with no waitlist and no staged rollout. It's available now in the Grok app, Cursor, Grok Build, and the xAI API. You can try it today.
Elon Musk framed the release as a "strong combination of intelligence, speed & low cost," and the technical details back up the speed and cost half of that claim. The model name is grok-4.7, the context window is 500,000 tokens, and the knowledge cutoff sits at May 2026. It handles text and image input and produces text output, with reasoning effort settings at low, medium, high, and xhigh.
If you use the Grok app on your phone, the model is simply the new default. Nothing to configure.
What Changed Under the Hood in Grok 4.7
xAI lists four changes from Grok 4.6, and together they explain where the gains come from.
The first is a new, larger base model. Grok 4.7 doesn't reuse the 4.6 base at all. That's a bigger structural change than a point release usually implies.
The second is a longer reinforcement learning run on harder tasks, with the training mix weighted toward problems that take many hours to complete. That's aimed squarely at agentic and coding workloads rather than quick question-answering.
The third is better self-verification and long-context handling. The company says the model checks its own work more carefully, which matters most on tasks where an early mistake compounds.
The fourth is native support for Grok Bot, xAI's agent framework. Grok 4.7 was trained to understand that tooling for conversational and knowledge work, which is a sign of how much of xAI's roadmap now runs through agent-style interactions.
Grok 4.7 Benchmarks
The launch table compares Grok 4.7 at xhigh effort against Grok 4.6 at high effort, plus two frontier competitors, GPT-5.6 Sol Max and Fable 5.1 Max. All scores are vendor-reported, so treat them as directional rather than independently verified.
| Benchmark | Grok 4.7 | Grok 4.6 | GPT-5.6 Sol Max | Fable 5.1 Max |
|---|---|---|---|---|
| CursorBench 4.0 | 46.3% | 40.4% | 41.7% | 51.8% |
| DeepSWE v1.1 | 71.0% | 65.2% | 72.7% | 70.0% |
| EEBench | 64.0% | 53.0% | 39.4% | 56.4% |
| Terminal-Bench 4.0 | 38.0% | 20.3% | 37.3% | 57.9% |
| HealthBench Professional | 56.7% | 48.5% | 60.5% | 62.1% |
Grok 4.7 improves on Grok 4.6 in every row. The largest jump is on Terminal-Bench 4.0, which nearly doubles from 20.3% to 38.0%. EEBench rises 11 points to 64.0%. That's the top score in the table.
The honest read, though, is that Grok 4.7 doesn't lead across the board. Fable 5.1 Max tops four of the seven benchmarks, including a 57.9% Terminal-Bench score, and GPT-5.6 Sol Max holds the best DeepSWE result at 72.7%. xAI's claim of a frontier price-performance position rests on cost, not on sweeping every metric.
Grok 4.7 vs Grok 4.6: Price, Speed, and Context
The comparison that matters most for most users is against the previous version, and here the numbers are simple. Both models start at $2 per million input tokens and $6 per million output tokens. The 500,000-token context window is identical. Speed is the same, with Grok 4.7 Fast offering the same model on faster infrastructure at twice the output speed and twice the price.
So the upgrade is free on paper. You get a larger base model, a longer RL run, and better self-verification for the same rates. The catch? Token consumption, retries, and tool calls still determine the real bill. A "same price" upgrade can cost more in practice if the new model runs longer reasoning chains.
One practical warning worth repeating from xAI's own docs: with a 500,000-token context window on both models, it's easy to walk off a pricing cliff. A 210,000-token request costs double what a 199,000-token request does for just 11,000 extra tokens. Setting a prompt cache key, as xAI recommends, helps keep cache hits reliable and reduces that exposure.
Grok 4.7 also ships with an entirely new safeguard stack. xAI calls it the strongest model it has tested on refusals and jailbreak resistance. It topped LatchBio's biosafety benchmark at 62.4%, and on HackerBench v0.3, xAI's own benchmark for risky and malicious cyber tasks, the model let 3.3% of risky dual-use prompts through. The company says it rarely blocks legitimate security work, and select cybersecurity partners now get invite-only access to its red-team capabilities for defense research.
On privacy, the usual cautions apply to any hosted model. Prompts you send through the Grok app or API are processed on xAI's servers, so avoid feeding it anything you wouldn't want stored elsewhere, and check your account settings for data-sharing controls. The Grok app requests the standard permissions an AI assistant needs; review them on install and decline anything unrelated to chatting and voice input.
What Grok 4.7 Means for Mobile Users
If you use Grok through the app rather than the API, the update is mostly invisible. The model you talk to got better at long, multi-step requests, which shows up in improved handling of tasks like research summaries or coding questions that need several steps of reasoning.
The Grok 4.7 Fast variant runs only in Cursor and Grok Build, not on the public xAI API, and it's excluded from Grok Build's free tier. For app users, the standard model is what you get, and it's still the same price as before.
There's also a US regional endpoint at us.api.x.ai/v1 that keeps inference in the United States at a 10% premium. If data residency matters to you, that's the option to know about, though it's aimed at developers rather than casual app users.
Should You Switch to Grok 4.7?
If you're on Grok 4.6 and paying for tokens, the upgrade is a no-brainer. Better results at the same headline price. No migration cost. Developer tools like Cursor treat it as a drop-in.
If you're on a competing frontier model, Grok 4.7 won't win you over on raw benchmarks alone. It trails Fable 5.1 Max and GPT-5.6 Sol Max on several measures. Where it stands out is coding and agentic work at a lower price. That's a specific case, not a universal one.
The Bottom Line on Grok 4.7
Grok 4.7 is a solid, same-price upgrade that targets coding and agentic tasks, and the Terminal-Bench jump is the headline number worth remembering. It doesn't dominate the frontier, and xAI's own launch table is honest about that. What it does is deliver clear gains over Grok 4.6 without asking for more money.
Download the Grok app from APKPure to try the new model on your phone, and if you already run the app, just open it. The update is automatic. For developers, the API cost is unchanged, so there's no reason to delay a switch unless your workflow depends on a competing model's specific strengths. Expect the next release in roughly a month, given xAI's current pace.




