LLMs

Grok 4.7 vs GPT-6.2 Sol: Same Input Price, Different Bill

Both charge $2 per million input tokens. Everything after that diverges.

🕐 7 min read 👁 27 views 📅 Nov 20, 2026
The short version
● Same input price Both charge $2 per million input tokens.
● Output Grok 4.7 is $6. GPT-6.2 Sol is $10. Grok wins on generation.
● Cached input Sol is $0.10. Grok is $0.50. Sol wins five to one on repeat context.
● Context Grok 4.7 holds 500k. Grok prices change above 200k prompt tokens.

These two land at the same headline input price and diverge everywhere else. Which one is cheaper depends entirely on the shape of your workload, and the deciding factor is not output length — it is how much of your prompt repeats.

The pricing, per million tokens

Grok 4.7GPT-6.2 Sol
Input$2.00$2.00
Cached input$0.50$0.10
Output$6.00$10.00
Above 200k prompt tokens$4 / $1 / $12No stated tier change
Context window500kNot stated
Model IDgrok-4.7gpt-6.2-sol

When Grok 4.7 is cheaper

Generation-heavy work. If your prompts are short and your outputs are long — drafting, long-form writing, code generation from a brief spec — the output price dominates and $6 against $10 is a 40% saving on the line item that actually moves. Grok also gives you an explicit reasoning-effort control with four levels, low through xhigh, so you can buy less thinking on easy calls.

When GPT-6.2 Sol is cheaper

Anything with a large, stable prefix. Cached input at $0.10 against $0.50 is a five-to-one difference, and in a retrieval or agent loop that resends the same system prompt, tool definitions and document context on every turn, cached tokens are the majority of what you pay for. That gap swamps the output difference quickly.

Sol also takes image input and returns text, and it is built for tasks that mix text, images and tools such as web search. If your pipeline calls tools, that is the model it was designed for.

The 200k cliff

Grok 4.7 doubles its rates above 200,000 prompt tokens: $4 input, $1 cached and $12 output. If you are feeding it whole codebases or long document sets, model the bill at the tier you will actually sit in rather than the headline. A workload that averages 250k prompt tokens is not a $2 workload.

A note on Grok 4.7 Fast

xAI also lists Grok 4.7 Fast, which it describes as the same model at twice the standard token rates, or 1.5x for long context. It is available in Cursor and Grok Build rather than the public xAI API, so it does not belong in a like-for-like API comparison — but it is worth knowing it exists before you conclude a price looks wrong.

For getting started with either, our complete Grok guide covers the account and API setup side.

FAQ

Is Grok 4.7 cheaper than GPT-6.2 Sol?
On output, yes: $6 against $10 per million tokens. On cached input, no: $0.50 against $0.10. Input is $2 on both.
What is Grok 4.7's context window?
500,000 tokens. Rates change above 200,000 prompt tokens, to $4 input, $1 cached and $12 output per million.
Which model is better for an agent loop?
GPT-6.2 Sol, in most cases. Agent loops resend a large stable prefix every turn, and Sol's $0.10 cached input is five times cheaper than Grok's $0.50.
Does GPT-6.2 Sol accept images?
Yes. It takes text and image input and returns text, and supports Batch and Flex.
⚖ Our Verdict

Generation-heavy work with short prompts goes to Grok 4.7 on its $6 output. Agent and retrieval loops with a large repeated prefix go to GPT-6.2 Sol on its $0.10 cached input.