Grok 4.6 Release Date: Official Launch and Availability
grok-4.6, the context window is 500,000 tokens, and the model is available through the xAI API and named gateway partners. EvoLink also lists a live Grok 4.6 route with current pricing. Developers looking for access, model ID, or price should use the Grok 4.6 API page; this article preserves the release timeline and explains what changed.The direct answer
| Question | Confirmed answer on August 13, 2026 |
|---|---|
| Is Grok 4.6 released? | Yes. xAI announced it on August 12, 2026. |
| What is the API model ID? | grok-4.6 — not the URL-style grok-4-6. |
| Is it available on EvoLink? | EvoLink lists the route as live with a current pricing surface. |
| What is the context window? | 500,000 tokens. |
| Which APIs are documented? | Chat Completions and Responses. |
| What should production teams do first? | Run representative tasks, validate usage and billing, and keep a tested fallback. |
The launch resolves the biggest pre-release uncertainties: model identity, API availability, context, pricing, reasoning controls, and partner access are now documented. Parameter-count rumors and other pre-release speculation are no longer useful for deciding how to integrate the model and have been removed from this page.
What xAI officially launched
low, medium, high, and xhigh.The standard direct-API rates for prompts below 200K tokens are $2 per million input tokens, $0.50 per million cached input tokens, and $6 per million output tokens. When total prompt input reaches 200K tokens, xAI applies the long-context tier: $4 input, $1 cached input, and $12 output per million tokens. Channel prices can differ, so always use the price surface for the route you will actually deploy.
| Launch fact | Official state | Why it matters |
|---|---|---|
| Model ID | grok-4.6 | Prevents requests from using the hyphenated page slug by mistake. |
| Context | 500K tokens | Supports large repositories and document sets, but also introduces long-context pricing. |
| Reasoning | low, medium, high, xhigh | Lets teams trade response time and usage against task difficulty. |
| API surfaces | Chat Completions and Responses | Supports conventional chat flows and tool-driven agent workflows. |
| Modalities | Text and image input; text output | Enables visual analysis without implying image generation. |
| Rate limits shown in the model docs | 150 RPS and 50M TPM | Useful as an upstream reference; effective limits still depend on account and route. |
Official availability and EvoLink availability are different facts
A provider launch proves that an upstream model exists. A gateway listing proves that a specific access layer has configured a route and commercial surface. Neither alone guarantees that every account, region, workload, or quota behaves identically.
grok-4.6, review the live price module, and test both Chat Completions and Responses if your workload depends on tools. EvoLink's value is that teams can evaluate Grok beside other models through one gateway and retain a fallback without rebuilding provider-specific authentication for every comparison.
What developers should verify before production
| Check | Minimum evidence | Stop condition |
|---|---|---|
| Model identity | Requested and returned model identity are consistent | Unexpected alias or route behavior |
| Cost | Usage fields reconcile with the EvoLink price surface | Unexplained token or tool charges |
| Long context | A task near and above 200K prompt tokens behaves as expected | Cost jump or quality regression is not understood |
| Reasoning | The selected effort level improves accepted-task quality | Higher effort only adds latency or output cost |
| Tools | Schema validity, tool arguments, recovery, and loop count pass | Repeated or malformed calls |
| Fallback | A second model can accept the shared request subset | Rollback requires an application rewrite |
FAQ
When was Grok 4.6 released?
xAI officially announced Grok 4.6 on August 12, 2026.
What is the Grok 4.6 API model ID?
grok-4.6. The hyphenated grok-4-6 form is used in page URLs and should not be sent as the API model value.Is Grok 4.6 available through EvoLink?
Yes. EvoLink lists a live Grok 4.6 route and current pricing. Check the model page for the active commercial surface before deployment.
How large is the context window?
The documented context window is 500,000 tokens. Long-context pricing begins when prompt input reaches 200,000 tokens.
Is Grok 4.6 the same price as Grok 4.5?
Their standard direct input and output rates are the same, but their cached-input rates differ. Gateway prices should be checked on the relevant live route.
Should teams replace Grok 4.5 immediately?
No automatic replacement is recommended. Replay real tasks, compare accepted-task cost and reliability, canary the winning workloads, and keep a rollback route.
