Grok 4.7 vs Grok 4.3

The flagship and the workhorse

Both models in this comparison think before they answer, both read images, and both are in our free chat. Grok 4.7 is xAI's newest flagship from September 2026, built on a new and larger base model. Grok 4.3 is the cheaper reasoning model from April 2026. On our site the main practical difference is the allowance: Grok 4.7 is limited to 10 messages a day, while Grok 4.3 Thinking can use the whole daily allowance of 30. So the question is simple: when is a message worth spending on 4.7?

Specs side by side

Prices: OpenRouter list price, Oct 2026. On xAI's own API, Grok 4.7 costs more for prompts above 200,000 tokens.

What is new in Grok 4.7

Grok 4.5 and 4.6 came between the two; Grok 4.7 then moved to a new, bigger base model. xAI presented it with a focus on coding, long agent tasks and following instructions more closely, together with updated safeguards. Its training data is about five months more recent than Grok 4.3's, so it knows about software releases and events from early 2026 that Grok 4.3 has never seen. We do not reprint xAI's benchmark charts; they were run at high effort settings that differ from how we run the model.

Grok 4.3 remains a capable reasoning model. For most everyday reasoning, such as checking a calculation, working through a plan or explaining a concept step by step, it does the job well and much more cheaply.

Speed in our chat

We run both at low reasoning effort and hide the reasoning text, so both show a short "Thinking…" pause before the answer. In our tests the waits were similar, a few seconds for a short question. Long, complex answers take longer on either model. If you just want an instant reply, Grok 4.20 is the fastest choice.

How to split your messages

A good habit is to start with Grok 4.3 Thinking and switch to Grok 4.7 when the task is hard or recent knowledge matters. Some examples:

• Grok 4.3: a study schedule, a budget check, explaining a formula, comparing two phone plans, reviewing a short email.
• Grok 4.7: reviewing a tricky piece of code, a long multi-part analysis, a question about something from 2026, or a second opinion when Grok 4.3's answer seemed off.

Writing a clear prompt helps either model, but it matters most with your limited Grok 4.7 messages. Put the whole task, the context and the format you want in a single message instead of spreading it over several short ones, and paste any code or text it needs to see. That way one Grok 4.7 message does the work of three.

When your 10 Grok 4.7 messages for the day are used up, our chat switches you to Grok 4.20, and you can pick Grok 4.3 Thinking from the model menu.

Cost for developers

Through the API, Grok 4.7 costs 1.6 times as much as Grok 4.3 for input and 2.4 times as much for output. A reply of 1,000 output tokens costs $0.006 on Grok 4.7 and $0.0025 on Grok 4.3 at list price, before counting reasoning tokens, which are billed as output too. On xAI's own API, Grok 4.7 prompts longer than 200,000 tokens are charged at double rates, $4 / $12, and cached input costs $0.50 per million tokens.

Defaults also differ. Grok 4.7 starts at high reasoning effort unless you lower it, while Grok 4.3 starts at low, so an untouched request to Grok 4.7 can produce far more hidden reasoning than one to Grok 4.3. If you switch an app from one to the other, set the effort explicitly and watch the token counts for a day. A faster variant, Grok 4.7 Fast, exists, but xAI offers it only in Cursor and Grok Build, not in the public API. On OpenRouter the ids are x-ai/grok-4.7 and x-ai/grok-4.3; our Grok API guide shows working requests.

What they have in common

Under the differences, the two behave much alike in daily use. Both follow the same Grok style, direct and fairly informal unless you ask otherwise. Both can misremember facts and sound sure of themselves, so check names, figures and quotes before you use them. And in our chat both have the same limits: no web or X search, no image generation, and only the recent part of a long conversation in view.

Choose Grok 4.7 if…

• the task is genuinely hard: complex code, long analysis, multi-step instructions;
• you need knowledge from the first half of 2026;
• you want the strongest Grok model and can live with 10 messages a day;
• you are a developer who needs the best quality and accepts a higher per-token price.

Chat with Grok 4.7

Choose Grok 4.3 if…

• you need careful step-by-step answers many times a day;
• your question does not depend on events after 2025;
• you paste very long text, where its one-million-token window helps;
• you are building something at scale and output price matters.

Chat with Grok 4.3 Thinking. More detail on the model pages for Grok 4.7 and Grok 4.3, or compare the quick model in Grok 4.20 vs Grok 4.3.

FAQ

Is Grok 4.7 always better than Grok 4.3?

It is the stronger model overall, but for many everyday reasoning tasks Grok 4.3 gives an equally good answer and costs far less.

Why is Grok 4.7 limited to 10 messages a day here?

It is xAI's most expensive model in our chat. The limit keeps it free for everyone.

Which knows about more recent events?

Grok 4.7, with a knowledge cut-off in May 2026, against December 2025 for Grok 4.3. Neither browses the web in our chat.

Which has the larger context window?

Grok 4.3, with one million tokens. Grok 4.7 has 500,000.