Grok 4.7 API
Grok 4.7 from xAI handles long, hard tasks and checks its own work as it codes, researches and analyzes within a 500K-token context.
Prices are temporarily unavailable — refresh the page in a minute.
What Grok 4.7 is good for
Grok 4.7 is xAI's model for long-running work, trained on a tougher mix of problems that take many steps to finish and better at reviewing its own intermediate results. It holds up to 500,000 tokens of context and offers four reasoning effort levels, from low to xhigh. It suits coding assistants, repository analysis, research and document-heavy professional work in fields such as law and engineering.
- Debugging and repository analysis across large codebases
- Multi-step research where the model double-checks its own findings
- Reviewing long documents, contracts and specs in one session
- Reports, briefs and structured analyses for office-style work
Grok 4.7 pricing
| Option | Price, USD | Video Blade tokens |
|---|---|---|
| Cached Input | tokens | |
| Input | tokens | |
| Output | tokens |
Prices are per 1M text tokens (about 750,000 words of model text) and live from our price list. Tokens are held when a run starts and settled to what was actually used; a failed run costs nothing.
Grok 4.7 is coming soon to Video Blade
It is listed with its live price so you can compare it. You can't order it through Video Blade yet.
Grok 4.7 FAQ
How much does Grok 4.7 cost?
1M text tokens from . Prices are live from our price list, and you see the exact price before anything runs.
What changed between Grok 4.6 and Grok 4.7?
xAI moved Grok 4.7 to a new, bigger base model and trained it longer with reinforcement learning, weighting that training toward difficult, long-running problems. It also verifies its own work more carefully and keeps track of long context better, and official benchmarks put it ahead of Grok 4.6 on all seven tests compared.
How does Grok 4.7 score on coding and engineering benchmarks?
Published results list 46.3% on CursorBench 4.0, 71.0% on DeepSWE v1.1 at high effort and 64.0% on EEBench, an electrical engineering test. The three cover different ground: software engineering, longer multi-step coding work and technical engineering knowledge.
Can I use Grok 4.7 on Video Blade today?
Not yet. Grok 4.7 is listed with its live price so you can compare it with the models you can use today; ordering it through Video Blade is coming soon.
How is Grok 4.7 billed?
Per million text tokens of the model, with separate rates for the text you send and the text you get back. Tokens are held when a run starts and settled to what was actually used; a failed run costs nothing. Current prices are shown on this page.
Grok 4.7 in the Video Blade API
grok-grok-4-7Ordering this model through the API is coming soon.