Grok 4.7
Low token price, long context. Text and image reasoning with a 500K-token window, function calling and structured outputs.
Which reasoning effort setting?
controls how long Grok 4.7 thinks before answering. Token prices never change; higher settings spend more and raise task cost.
Start at High. It scores 35.2 against 35.5 at the top setting, for 20% less per benchmark task. Rule: cheapest setting within 2 points of the model’s best score.
At High, Grok 4.7 thinks for about 58.2K tokens per task, then writes about 30.9K more as its answer: $6.18 per task, 2.1× Low, for +4.1 points.
Output rate: $6.00/1M at every effort level. Thinking tokens are billed as output.
| Effort | Index | Per task | Thinking | All output | × Low | Time |
|---|---|---|---|---|---|---|
| Low | 31.1 | $2.92 | 21.7K | 42.7K | 1.00× | 10.3 min |
| HighStart here | 35.2 | $6.18 | 58.2K | 89.1K | 2.12× | 20.5 min |
| Xhigh | 35.5 | $7.77 | 69.4K | 105K | 2.66× | 24.9 min |
Per task of the Artificial Analysis Intelligence Index: the average over its nine evaluations, counting input, cache, and output tokens at list prices. These are long, agent-style benchmark tasks, much longer than a chat message.
Where Grok 4.7 sits on the frontier
Intelligence Index against one 100K-input, 10K-output request.