← All articles

Grok 4.7: xAI's newest flagship is already on the site

A glowing abstract wave of blue and white data streams — a metaphor for a new flagship model version

On September 22 our provider kie.ai opened access to Grok 4.7 — xAI's newest flagship model. It is already available on NeuralSpace: in chat, in the Code section, and through the public API. Here is what is known about it, how it differs from the previous flagship Grok 4.6, and how to try it.

What the model is

Grok 4.7 is the next version of the Grok family after 4.6, which has been our flagship for code and complex tasks. According to the provider's announcement, the model brings enhanced reasoning for complex problem-solving and analysis, improved coding performance — development, debugging, technical workflows — and better understanding of complex multi-step instructions. Conversational intelligence is highlighted separately: the model should handle dialogue more naturally.

There were no independent benchmark numbers in the announcement, so it is too early to compare versions by figures — but the flagship handover follows the usual pattern: every new Grok 4.x version raised the bar in code and agentic tasks, and 4.7 follows the same path.

Branching glowing lines converge into one bright point — a metaphor for reasoning over a complex task

What it means in practice

Like Grok 4.5/4.6, the new model runs on the provider's Responses API protocol and supports an adjustable reasoning level: run a simple question cheaply and a complex chain at maximum, without changing the model. In the site's chat, Grok 4.7 has the reasoning level selector and web search, like the rest of the family.

For code, the model is available in the Code section — it handles multi-step work: project edits, debugging, tool calls. For automations, the model is included in the public API and MCP: the identifier is grok-4-7, tool calling is supported, so it can be used in agentic scenarios just like Grok 4.6.

A glowing data stream passes through a grid of identical cells — a metaphor for one API for many models

Pricing

Grok 4.7 pricing on NeuralSpace matches Grok 4.5/4.6: $0.8 per 1M input tokens and $2.4 per 1M output tokens — before the Code section markup. The minimum balance to start a conversation is the same as for the rest of the family. Switching from 4.6 to 4.7 changes nothing in price — only in answer quality.

How to try it

Pick Grok 4.7 in the model list of the chat or the Code section — or pass grok-4-7 in the public API. The older models in the family (Grok 4.3/4.5/4.6) remain available: if a task is better solved by a proven version, there is no need to switch.