Grok 4.7 vs Grok 4.6

API price, context window and output limit for Grok 4.7 and Grok 4.6 from Kingy's checked data, and what common workloads cost on each. Every price shows when Kingy last checked it.

At a glance

Verdict: Both cost the same per token. In Kingy's tests, Grok 4.7 gave fewer wrong answers on the AI Accuracy Index (0.4% against 4.8%) with a similar share right; Grok 4.6 passed 20/20 on Kingy's coding and data tasks, Grok 4.7 19/20.

Same standard price: $2 in / $6 out per 1M tokens.
Same context window: 500K tokens.
Both charge a higher rate when one request's input is over 200K tokens.
Grok 4.7's price was read automatically from the official page by fixed rules, not reviewed by a person.
Grok 4.6's price was read automatically from the official page by fixed rules, not reviewed by a person.

Measure	Grok 4.7	Grok 4.6
Provider	xAI	xAI
API price	$2 in / $6 outStandard, under 200K prompt tokensCached input $0.5$4 in / $12 outStandard, 200K prompt tokens or more (whole request billed at this rate)per 1M tokens, USD	$2 in / $6 outStandard, under 200K prompt tokensCached input $0.5$4 in / $12 outStandard, 200K prompt tokens or more (whole request billed at this rate)per 1M tokens, USD
Context window	500K tokens	500K tokens
Max output	Not listed	Not listed
Availability	—	—
Retirement	None listed in Kingy AI Deadlines	None listed in Kingy AI Deadlines
Price check	Read automatically, not human-reviewedChecked Nov 12, 23:23 UTC	Read automatically, not human-reviewedChecked Nov 12, 23:23 UTC
Official sources	docs.x.ai › models	docs.x.ai › models

Larger is not always better: a bigger context window or output limit only matters if your work needs it. These figures compare price and limits, not answer quality.

What common workloads cost

At 10,000 requests a month, standard list rates. Change the numbers in the tool below for your own usage.

Measure	Grok 4.7	Grok 4.6
Chat reply2,000 in / 500 out	$70.00 a month$0.007 per request	$70.00 a month$0.007 per request
Coding agent step30,000 in / 2,000 out	$720 a month$0.07 per request	$720 a month$0.07 per request
Long document summary150,000 in / 1,500 out	$3,090 a month$0.31 per request	$3,090 a month$0.31 per request
Classify short text300 in / 20 out	$7.20 a month$0.00072 per request	$7.20 a month$0.00072 per request

Excludes prompt-cache discounts, batch and priority pricing, regional rates, tool fees and tax. Reasoning tokens count as output, so a model that thinks longer costs more per request than these figures show.

What Kingy measured

Kingy's own runs, not vendor claims. These are small tests: read each one for its tasks and limits before generalising. All results are also on the AI Model Board.

Test	Grok 4.7	Grok 4.6
AI Accuracy Index: 125 official figures from memory, 2 asks each	Right 19%, wrong 0.4%, not sure 78%2026-11-10 · read the test	Right 22%, wrong 4.8%, not sure 74%2026-11-12 · read the test
Kingy's four coding and data tasks, 5 tries each	19/20 passed · Kingy's bill $0.135722026-11-12 · read the test	20/20 passed · Kingy's bill $0.087602026-11-12 · read the test

Compare other models

First model Second model

Loading Kingy's price data…

Side by side

Measure

What would it cost you?

Input tokens per request Output tokens per request Requests per month

Standard list rates only. Excludes prompt-cache discounts, batch and priority pricing, regional rates, tool fees and tax. Reasoning tokens count as output. A long-context rate applies when one request's input passes the model's threshold.

Prices of other AI models

See the checked prices of all the main AI models in one table, or pick any two models in the tool above.

How this works

Prices come from the same checked data behind Kingy's AI cost calculator and the Kingy Facts API. OpenAI, Anthropic and Google prices are reviewed by a person and re-checked against the official pricing page every few hours. Prices for other providers are read automatically from their official pricing pages by fixed rules, and the table says so.

"Not confirmed right now" means Kingy could not confirm the price within its check window, so it shows no number. Use the official link instead. Retirement dates come from Kingy AI Deadlines. Every source re-check is logged in the public accuracy ledger.

Why isn't my model listed?

Only models with a tracked official price page are listed. Embedding, rerank and moderation models are left out because they have no output price to compare.

Does this show benchmarks?

It shows Kingy's own tests when a model has been through them: the AI Accuracy Index and Kingy's coding and data tasks. These are small tests, so read each one for its tasks and limits before generalising. Vendor benchmarks are left out, because results depend on the test and settings used; Kingy's full comparisons cover them when one exists for your pair.