Grok 4.7: price, features and how it is used
Grok 4.7 is xAI’s leading model, described as suitable for code and agentic tool calling with configurable reasoning. It was announced on September 21, 2026. Its knowledge extends to May 2026.
Provider: xAI · Released: September 2026
Results by test
| Test | Result | Source |
|---|---|---|
| CursorBench 4.0 (xhigh effort) | 46,3% | xAI, Grok 4.7 announcement |
| DeepSWE v1.1 (high effort) | 71% | xAI, Grok 4.7 announcement |
| EEBench (xhigh effort) | 64% | xAI, Grok 4.7 announcement |
| Terminal-Bench 4.0 (xhigh effort) | 37,6% | xAI, Grok 4.7 announcement |
| HealthBench Professional (xhigh effort) | 56,7% | xAI, Grok 4.7 announcement |
| Harvey Legal Agent Benchmark (xhigh effort) | 19,6% | xAI, Grok 4.7 announcement |
Intelligence, price and speed by effort level
There is no data yet for this model on effort level, intelligence and price per task.
Data on the Intelligence Index, capability indexes, price per task, speed and latency: Artificial Analysis. Prices per 1M tokens are from official vendor data.
Compare Grok 4.7 with another model
Choose a model and see prices, context and results side by side.
How to use Grok 4.7
Grok 4.7 is available through the Grok API, in Cursor and in Grok Build (x.ai/build), as well as through external coding tools, model routers and cloud platforms. Access through grok.com, the X app or subscription plans is not described in the official pages.
- Get a Grok API key from xAI. If you want to try it without code, open Cursor or Grok Build.
- Use the Responses API with the grok-4.7 model. Chat Completions is marked as legacy and has a link to a migration guide.
- The effort level is set with the reasoning_effort parameter. The allowed values are not listed on the models page, so check them in the reasoning documentation.
- In the Responses API the model always returns reasoning.encrypted_content, even if you did not request it in include.
- For non-English text, keep in mind that non-English text uses more tokens than English, so the same text costs more. Check the quality of answers on your own real examples.
What it is suited for
- Price $2 input and $6 output per 1M tokens with a 500K token context window.
- Configurable reasoning via the reasoning_effort parameter.
What to keep in mind
- Reasoning_effort levels are set by the user. xAI does not publish per-level figures for tokens, price and latency.
- The comparisons in the announcement are by the vendor itself. In the published table, Fable 5.1 is ahead of Grok 4.7 on CursorBench 4.0, Terminal-Bench 4.0 and HealthBench Professional.
- The pricing page does not describe price levels by request size.
Frequently asked questions about Grok 4.7
- How much does Grok 4.7 cost?
- USD 2 per 1M input and USD 6 per 1M output tokens, at a context of 500K tokens. There is also a fast variant with twice the output speed at twice the price.
- Is Grok 4.7 free?
- Partly. xAI offers a free trial in Grok Build, but the limits are not specified. The API is billed per token.
- How does Grok 4.7 differ from Grok 4.6?
- According to xAI it is offered at the same price and with the same speed as Grok 4.6. A more detailed comparison is not published on the xAI pages that the sources link to.
- How do you use Grok 4.7 in other languages?
- Write in your own language via the API or in Cursor and check the answers against your own examples. Non-English text uses more tokens, so the same request will cost more than in English.
More models in AI Benchmark
- Gemini 4 Neon (Google)
- Gemini 4 Carbon (Google)
- GPT-6 Astra (OpenAI)
- GPT-6.1 Sol (OpenAI)
- GPT-6.1 Terra (OpenAI)
- Muse Wonder Full (Meta)
Intelligence Index, token, TTFT and effort are explained in the AI Benchmark glossary. Prices are per 1M tokens and may change, so check the provider’s official page.
Sources: xAI, model documentation and prices · xAI, Grok 4.7 announcement
Last checked: 18.11.2026
Compare with other models
Set Grok 4.7: price, features and how it is used to another model by price, context and benchmarks.