NewsxAI1 min read
Grok 4.7: xAI's New Model for Coding and Agents
Grok 4.7 reaches the API, Cursor and Grok Build at $2 / $6 per million tokens, with a 500K-token context. What it offers and how the Grok family looks.
Key points
- Grok 4.7 shipped on 21 September as xAI's most capable model for code and professional work.
- It costs $2 per million input tokens and $6 output, doubling above 200K tokens.
- It has a 500K-token context and is available in the xAI API, Cursor and Grok Build.
- For simpler tasks, Grok 4.3 costs less and reaches a 1M-token context.
Grok 4.7 was released on 21 September as xAI’s most capable model for coding and long professional work. xAI, Elon Musk’s AI company, is now part of SpaceX, which also owns Cursor. That is why the model arrived at the same time in the API, in Cursor and in Grok Build, xAI’s coding tool.
What improves
According to xAI, Grok 4.7 starts from a larger base model than Grok 4.6 and was trained more on hard, multi-hour tasks, so it checks its own work better and copes with long contexts. The company reports these results:
| Benchmark | Result according to xAI |
|---|---|
| CursorBench 4.0 (coding in Cursor) | 46.3% |
| DeepSWE v1.1 (real-world software engineering) | 71.0% |
As always with a company’s own numbers, wait for independent evaluations. For reference, Google reports 77.9% on DeepSWE v1.1 for Gemini 4 Argon.
Price and the Grok family
| Model | Context | Input / output (per million) |
|---|---|---|
| Grok 4.7 | 500,000 | $2 / $6 |
| Grok 4.3 | 1,000,000 | $1.25 / $2.50 |
| Grok Build 0.1 | 256,000 | $1 / $2 |
Prices double when the prompt goes over 200,000 tokens, and cached input is much cheaper. Grok 4.7 is noticeably cheaper on output than Anthropic’s or OpenAI’s top models, so it can pay off for agents that write a lot.
What changes for you
- If you use Cursor, you can already pick it as your model and compare it with Claude or GPT on your own tasks.
- If you build software, it is in the xAI API. See it next to the rest in our model comparison.
