Guides / Language models

Grok 4.7 explained: price, 500K context and where it fits

Updated November 2, 2026. Researched and drafted with AI help; every figure is checked against the linked source.

xAI released Grok 4.7 on September 21, 2026, and calls it its most powerful model for coding and knowledge work. Its headline price is unusual for a frontier model: $2 per million input tokens and $6 per million output tokens, less output price than most mid-tier models. This guide covers what the model is, what it costs on real workloads, and the three pricing details that change the picture: cache reads, the 200K-token tier and a batch discount smaller than other providers offer.

The short version

• Price: $2 input / $6 output per million tokens. Cached input: $0.50 per million.
• Long prompts: from 200,000 tokens, xAI’s pricing page lists $4 input, $1 cached input and $12 output.
• Context: 500,000 tokens. Modalities: text and image.
• API name: grok-4.7. Function calling, structured outputs and reasoning, with effort levels low, medium, high and xhigh (high is the default).
• Rate limits (xAI model page): 150 requests per second and 50,000,000 tokens per minute.
• Batch API: xAI’s pricing page now lists a 20% batch discount for Grok 4.7 ($1.60 input, $0.40 cached input, $4.80 output). Earlier versions of this guide said there was none. Anthropic, OpenAI and Google take 50% off for batch.
• Fast variant: xAI’s announcement says a fast variant gives twice the output speed at twice the price. The docs list no separate price for it, so it isn’t in our tables.

What changed from Grok 4.6

xAI says Grok 4.7 uses a new, larger base model than Grok 4.6, trained with a longer reinforcement learning run on harder and longer tasks, and that it works longer on difficult problems and checks its own work more carefully. The price didn’t move: xAI’s pricing page lists Grok 4.6 at the same $2 / $6 with $0.50 cached input, and both have a 500K context.

The announcement also includes xAI’s own benchmark results. These are the maker’s numbers, not independent tests. A few of them, Grok 4.7 against Grok 4.6:

In the same tables xAI shows Fable 5.1 (at its Max setting, against Grok 4.7 at xhigh) ahead of Grok 4.7 on CursorBench (51.8%) and Terminal-Bench (57.9%), and Grok 4.7 ahead on xAI’s EEBench (64.0% vs 56.4%). Vendor benchmarks pick the settings and the tasks, so treat them as a reason to test, not as a ranking.

What it costs on real workloads

We use the same three workloads as our cost per 1,000 requests guide, plus one output-heavy job. All figures are for 1,000 requests at official list prices, standard tier.

• Chat: 2,000 input + 500 output tokens.
• Agent step: 2,000 fresh input + 18,000 cached input + 1,000 output tokens.
• Summary: 10,000 input + 300 output tokens.
• Long writing: 2,000 input + 8,000 output tokens (a report or a long code file).

* DeepSeek peak-hour prices. Off-peak input and output cost half.

† Mistral Large 4 is in public preview. Mistral cut its price to $0.70 / $2.10 with $0.07 cache reads when it released the open weights in late October 2026 (it was $1.36 / $4.18 with $0.14 cache reads before).

Where Grok 4.7 fits

Cheapest of the $2-input models, unless you cache. For chat, summaries and anything output-heavy, Grok 4.7 costs less than Claude Sonnet 5.5, GPT-6.1 Sol, Gemini 3.1 Pro and GPT-5.6 Terra, because its $6 output price is 40% below the $10 of Sonnet 5.5 and GPT-6.1 Sol and half the $12 of Gemini 3.1 Pro and GPT-5.6 Terra. On long writing the gap is largest: $52 per 1,000 against $84 and $100.

Caching flips the agent column. Grok 4.7’s cached input is $0.50 per million, a 75% discount. Claude Sonnet 5.5 and GPT-6.1 Sol charge $0.10 (95% off). On a cache-heavy agent step that is enough to put Grok 4.7 behind both of them ($19.00 vs $15.80 per 1,000 steps), and roughly level with Gemini 3.1 Pro ($19.60). The more of each request is a repeated context, the more this matters. Our prompt caching guide has a document Q&A example where the gap is wider: $27.80 for Grok 4.7 against $9.40 for Sonnet 5.5 and GPT-6.1 Sol.

Much cheaper than other flagships. xAI positions Grok 4.7 as its frontier model. Against Claude Opus 5.5 it costs under half for chat, summaries and long writing, but 60% as much on cache-heavy agent steps ($19.00 vs $31.60), again because of the cache price. Against GPT-6 Astra it costs roughly 12% to 22% as much on these workloads (the high end is the agent step).

Not the cheapest overall. DeepSeek V4 Pro and Mistral Large 4 cost less than Grok 4.7 on every workload here, even at list price and DeepSeek’s peak hours. After its late-October price cut, Mistral Large 4 costs about a third of Grok 4.7 on chats, summaries and long writing ($2.45 vs $7.00 per 1,000 chats, $18.20 vs $52.00 per 1,000 long writes). Whether they do the job as well is something to test with your own prompts.

The 200K-token tier

Grok 4.7 accepts 500,000 tokens, but from 200,000 tokens xAI lists double prices: $4 input, $1 cached input and $12 output. Here is one request with a 300,000-token prompt and a 2,000-token answer, without caching:

At this size, input is almost the whole bill, so Grok 4.7’s cheap output stops helping. Anthropic says Claude 4.6 and later models include the full 1M context at standard pricing (Haiku 5.5 is the exception), which is why Sonnet 5.5 costs about half here. If your prompts often go past 200K tokens, that matters more than the headline price. One caveat for every Claude comparison: Anthropic says Claude 4.7 and later models use a tokenizer that produces about 30% more tokens for the same text, so the same document may not be 300,000 tokens on every model.

Tools and search cost extra

Grok has no real-time knowledge unless you turn on its search tools, and those are billed separately. xAI’s pricing page lists Web Search at $5 per 1,000 calls, X Search at $5 per 1,000 posts (and $10 per 1,000 profiles), and Code Execution at $5 per 1,000 calls. An agent that searches on every step can spend more on tool calls than on tokens: 1,000 steps with one web search each add $5 to the $19 token bill above.

Grok 4.7 in the Grok apps

The API and the consumer apps are priced separately. On October 12, 2026, xAI replaced the SuperGrok plans on its pricing page with The X Subscription: Plus at $8 a month, Premium at $30 (which xAI says includes the benefits of SuperGrok), Super at $100 and Ultra at $200. xAI says existing SuperGrok subscribers can keep their current plan and price. The announcement says Grok 4.7 is available in Cursor, Grok Build, the Grok API and through model routers and cloud platforms, but does not say which app plans include it. See subscription plans for the current prices of all the major chat plans.

Should you use it?

• Good fit: output-heavy work (long answers, code generation, reports) where you don’t reuse a big cached context, and prompts that stay under 200K tokens.
• Check alternatives: cache-heavy agents and repeated-document Q&A, where Claude Sonnet 5.5 and GPT-6.1 Sol are cheaper per step; very long prompts, where Sonnet 5.5’s flat pricing wins; and large offline jobs, since its 20% batch discount is smaller than the 50% Anthropic, OpenAI and Google offer.

Compare it head to head with Claude Sonnet 5.5, GPT-6.1 Sol or Gemini 3.1 Pro, or put your own token counts into the cost calculator. The Grok 4.7 page shows today’s community votes on whether it feels dumber than usual, next to official incidents, and you can embed a live rating with a badge.

Sources

1. xAI: Grok 4.7 announcement
2. xAI docs: Grok 4.7 model page
3. xAI docs: Models and pricing
4. xAI docs: Pricing
5. xAI: Introducing The X Subscription
6. xAI: Grok plans
7. Anthropic: Claude API pricing
8. OpenAI developer docs: pricing (GPT-6 models)
9. OpenAI developer docs: GPT-6.1 Sol model page
10. OpenAI API pricing
11. Google: Gemini API pricing
12. DeepSeek API pricing
13. Mistral docs: Mistral Large 4
14. Mistral: Mistral Large 4 open weights announcement
15. Mistral docs: API pricing