Grok 4.7
Grok 4.7 is SpaceXAI's September 21, 2026 frontier model with the API ID grok-4.7, a 500,000-token context window, text and image input, text output, no documented output-token limit, and low through xhigh reasoning effort.
Grok 4.7 replaces 4.6 as SpaceXAI's current frontier API model while keeping the same public token prices and context contract. It also separates the public API model from a Fast route available through Cursor and Grok Build, preventing an easy but costly availability assumption.
SpaceXAI's announcement, model page, and release notes control identity, limits, pricing, regional access, and Fast availability. Launch benchmarks and claims about hallucination, coding, or agent performance remain vendor-reported.
- Run coding and repository agents.
- Power native Grok Bot workflows.
- Analyze text and images in knowledge work.
- Use long context with explicit threshold-cost controls.
- Route between public API and product-only Fast access.
Use grok-4.7 for the public SpaceXAI API. It accepts text and images, returns text, supports a 500K context window, uses high reasoning by default, and always returns encrypted reasoning content on the Responses API.
Below 200K prompt tokens, rates are $2 input, $0.50 cached input, and $6 output per million. Above 200K, those rates double. Grok 4.7 Fast costs twice the applicable rate and is available through Cursor and Grok Build, not the public API.
SpaceXAI says 4.7 was trained to understand the native Grok Bot harness and positions it as the primary Grok Bot model. That is a product integration advantage, not proof that every external coding harness will see the same gains.
Grok 4.7 FAQ
Common questions about Grok 4.7.
What is the Grok 4.7 API model ID?+
Use grok-4.7. It supports text and image input, text output, a 500K context window, and low through xhigh reasoning effort.
How much does Grok 4.7 cost?+
Below 200K prompt tokens, rates are $2 input, $0.50 cached input, and $6 output per million. Above 200K, all three rates double for the request.
Can I use Grok 4.7 Fast through the public API?+
No. SpaceXAI documents the Fast variant through Cursor and Grok Build at twice the token rates, not on the public SpaceXAI API.