Grok app access
There is no official information showing Grok 4.7 is available in the Grok app. A subscription or model-picker screenshot would not mean the developer API is open.
As of September 18, 2026, Grok 4.7 cannot be called on EvoLink yet. Elon Musk has discussed the model and a delay, but xAI's model catalog, release notes and news page have no formal Grok 4.7 release entry. Use this page to track access and prepare an evaluation; keep an already tested model for work that needs to ship.
An announcement, an app rollout and a gateway API integration are separate milestones. None should be inferred from another.
Musk's public statements confirm that Grok 4.7 is being discussed and that an earlier target slipped. They do not set a final release date, and they do not mean the API has a documented way to call it.
There is no official information showing Grok 4.7 is available in the Grok app. A subscription or model-picker screenshot would not mean the developer API is open.
As of September 18, xAI's model catalog and release notes have no formal 4.7 entry. Official identity, supported controls and commercial terms still need confirmation.
This page does not provide a Grok 4.7 request example or price for now. Once the model ID, supported features and billing are confirmed, we will update EvoLink's integration status here.
We only update this status when official documentation or the actual integration changes. A provider announcement does not mean the API is usable, and this page will not switch to "live" because of one.
Leave your email for updates about Grok 4.7 access through EvoLink. This is an EvoLink notification list, not an xAI waitlist or a promise of early access.
Your email will be used for this EvoLink model update list. Joining does not guarantee access or a launch date.
Prefer community updates? Join EvoLink on Discord
You will know as soon as Grok 4.7 can be called on EvoLink. The notice includes what you can actually use: the real model ID, applicable pricing, supported request features and where to start testing.
These are suggested evaluation workloads, not confirmed Grok 4.7 capabilities. Start with a task your current model struggles to complete or costs too much to run.
Save a failing test, the repository revision and the allowed edit scope. Compare whether the candidate fixes the issue without unrelated changes, rather than judging the confidence of its explanation.
Prepare a sandbox with stable tool outputs and a strict action budget. Record repeated calls, recovery from errors and whether the final answer matches the work actually performed. Run only after tool support is confirmed.
Keep source documents and expected answers together. Score citations, schema validity and missed constraints separately. A published context limit will not establish that every long-document task is reliable.
Save current success rate, total task time and billed cost including retries. Set a workload-specific acceptance threshold before testing the candidate, and keep your working model configuration for rollback.
Pending does not mean free, unlimited or unsupported. It means that item has no official confirmation yet.
| Field | Current state | What your team should verify |
|---|---|---|
| Model ID | Not confirmed | Use the model ID published for the channel you will call. The page URL is not an API model ID. |
| Price and billing | Not confirmed | Check input, output, cache and tool charges on the channel you actually use, then reconcile a sample request with usage. |
| Context and output limits | Not confirmed | Confirm supported limits and any pricing thresholds. Do not inherit Grok 4.6 values. |
| Tools and structured output | Not confirmed | Test your tool schemas and output validator against the way of calling it stated in the official docs. |
| Modalities and reasoning controls | Not confirmed | Verify accepted inputs and controls before making them part of your product flow. |
| Account and regional access | Not confirmed | Confirm eligibility, effective limits and data requirements for your account and workload. |
A unified gateway can reduce account and integration overhead. The model still needs to earn its place with evidence from your application.
Replay representative Grok 4.6 tasks using the same acceptance rules. An improvement on hard code changes is useful only if required extraction, translation or latency behavior remains acceptable.
Compare completed-task cost and quality, then account for tool adaptation, prompt changes and reviewer effort. Keep Claude Opus 5 workloads that already meet your delivery requirements.
A successful request establishes a starting point. Before customer traffic, check returned identity, usage, error handling, effective limits and the request features your application actually needs.
Divide all evaluation charges by accepted outcomes. A low token price can still produce an expensive workflow when retries, excessive output or manual correction increase.
EvoLink already has product pages for Grok 4.6 and Claude Opus 5. Review their current integration details and run a small account-level test before committing production work. Neither is presented here as a measured equivalent to Grok 4.7.
Keep a known Grok workload and its acceptance results ready for a future paired upgrade test.
Use your current Claude workflow to estimate the benefit a second model would need to justify migration.
Use a shared integration surface where supported, while validating model-specific tools, limits and billing separately.
See the sourced pre-announcement, the delay and what would count as an actual release.
Build a regression set, define acceptance rules and keep a rollback plan.
Work through task fit, completed-task cost and the effort of adding another provider.
Check the existing EvoLink product page for its current integration and pricing details.
As of September 18, 2026, Grok 4.7 cannot be called on EvoLink yet. This page tracks its integration progress, and you can join the launch alert here.
The formal model ID is not confirmed yet. Do not turn the grok-4-7 in this page URL or a third-party SDK entry into an API request value. Go by the official documentation once it is published.
Grok 4.7 pricing has not been confirmed for EvoLink. Input, output, cache and tool charges each need to be confirmed. Grok 4.6 prices are not a Grok 4.7 quote.
No. A provider release and an EvoLink integration are separate events. Only after EvoLink finishes its integration and confirms supported features and billing will this page say the API is available.
No. It registers interest in EvoLink integration updates. It is not an official xAI waitlist, a reserved model slot or a promised launch date.
Those capabilities are not confirmed yet. Keep your schemas and validators, then run them against the official documentation once Grok 4.7 can be called.
Do not assume that consumer subscription access or message quotas establish developer API access, credits or limits. Check the terms and account requirements of the channel you plan to use.
Keep a model that already meets your delivery requirements. Prepare a representative evaluation and revisit the decision after Grok 4.7 identity, access, price and workload results can be checked.