Grok 4.7 is xAI's most capable model for coding and knowledge work, announced on September 21, 2026. It ships with a 500,000 token context window, configurable reasoning effort, and a public price of $2.00 per million input tokens and $6.00 per million output tokens. This guide covers the official numbers, the published benchmark table, and the five places you can actually reach the model.
The model profile is on Grok 4.7, and the wider picture is in Frontier AI Models 2026.
Grok 4.7 Pricing: The Official Numbers
xAI publishes Grok 4.7 pricing in its developer documentation. There is no separate free API tier for the model itself.
| Item | Official price |
|---|---|
| Input | $2.00 / 1M tokens |
| Output | $6.00 / 1M tokens |
| Context window | 500,000 tokens |
| Reasoning effort | Configurable |
Two separate Grok APIs cover other media. The Voice API prices speech-to-speech at $0.08 per minute, text-to-speech at $15.00 per million characters, and speech-to-text at $0.10 per hour for batch or $0.20 per hour for streaming. The Imagine API starts at $0.02 per image and $0.02 per second of video.
Cost per task matters more than cost per token for agentic work, because a longer reasoning trace multiplies output tokens. A run that reads 120,000 input tokens and produces 8,000 output tokens costs roughly $0.24 plus $0.05, so $0.29 per completed task.
Official Benchmark Table
The announcement page compares Grok 4.7 against Grok 4.6, GPT-5.6 Sol and Fable 5.1 on the same tasks, with prices shown alongside the scores.
| Benchmark | Grok 4.7 (xHigh) | Grok 4.6 (High) | GPT-5.6 Sol (Max) | Fable 5.1 (Max) |
|---|---|---|---|---|
| Input $ / 1M | $2 | $2 | $4 | $10 |
| Output $ / 1M | $6 | $6 | $20 | $50 |
| CursorBench 4.0 | 46.3% | 40.4% | 41.7% | 51.8% |
| DeepSWE v1.1 | 71.0% | 65.2% | 72.7% | 70.0% |
| EEBench | 64.0% | 53.0% | 39.4% | 56.4% |
| AA Briefcase v1.1 | 1,657 | 1,546 | 1,487 | 1,678 |
| Terminal-Bench 4.0 | 37.6% | 20.3% | 37.3% | 57.9% |
| Harvey Legal Agent | 19.6% | 15.8% | 2.5% | 6.7% |
| HealthBench Professional | 56.7% | 48.5% | 60.5% | 62.1% |
Read the table by column, not by row. Fable 5.1 leads on the terminal-heavy and legal benchmarks and costs five times more per output token. GPT-5.6 Sol leads on HealthBench Professional and costs more than double Grok 4.7 on output. Grok 4.7 leads on EEBench and on Terminal-Bench 4.0 it is effectively level with GPT-5.6 Sol while costing a third as much per output token, which is the strongest cost-per-task argument in the table.
One caveat worth stating plainly: these are vendor-published results on vendor-selected benchmarks. The asterisk on the DeepSWE figure marks a high-effort configuration. Treat the table as the official claim, not as an independent evaluation.
Where Grok 4.7 Is Available
- xAI API — the canonical endpoint, with the full model documentation at
https://docs.x.ai/developers/grok-4-7. - OpenRouter —
x-ai/grok-4.7lists 500,000 context tokens at $2 per million prompt tokens and $6 per million completion tokens, which confirms the official list price. - Cursor — selectable from the model list, and the CursorBench scores above come from Cursor's own harness.
- GitHub Copilot — available in Copilot since the September 21, 2026 release.
- Microsoft Foundry — announced for the Azure AI Foundry service.
If you need a second opinion on model choice for research work, Perplexity AI pairs a frontier-class model with live search and citations. For a self-hosted comparison, see Claude and the Frontier AI Models 2026 collection.
Practical Notes for Coding Agents
Grok 4.7 was trained with a longer reinforcement-learning run on harder, multi-hour tasks, and it is specifically noted as better at verifying its own work and managing long context. In practice that means two things for an agent workflow: it holds a large diff in context without losing track of it, and it tends to run more reasoning steps before it stops, which is why the output price matters more than the input price.
The native understanding of the Grok Bot harness also means it is the better fit for conversational and general knowledge tasks inside that harness.
Frequently Asked Questions
How much does Grok 4.7 cost per million tokens?
xAI lists $2.00 per million input tokens and $6.00 per million output tokens for Grok 4.7, with a 500,000 token context window. Voice and Imagine API calls are priced separately.
Does Grok 4.7 have a free API tier?
No. The Grok 4.7 API is metered. Chat access to the model inside the Grok product is a separate consumer subscription, and the developer API bills per token.
How large is the Grok 4.7 context window?
500,000 tokens. OpenRouter lists the same 500,000 token context for x-ai/grok-4.7.
Is Grok 4.7 cheaper than GPT-5.6 Sol or Fable 5.1?
On output tokens, yes: $6 per million against $20 for GPT-5.6 Sol and $50 for Fable 5.1. On the published benchmark table Grok 4.7 leads EEBench and matches GPT-5.6 Sol on Terminal-Bench 4.0 at a third of the output price.
Which tools support Grok 4.7?
The xAI API, OpenRouter, Cursor, GitHub Copilot and Microsoft Foundry all list it. Cursor publishes model documentation, and Copilot shipped it on September 21, 2026.
