SuperGrok vs ChatGPT Plus: 30 Days Tested, Now Updated for Grok 4.7

What Grok 4.7 actually changed

Served at the same price and speed as 4.6, on a new and larger base model with a longer reinforcement learning run weighted toward tasks that take hours. xAI's own figures, 4.7 at xHigh against 4.6 at High:

Terminal-Bench 4.0 nearly doubled at no extra cost, which is the single largest free upgrade either side of this comparison has shipped since June. It does not change the headline verdict — ChatGPT Plus still has Codex and Grok still has no dedicated coding agent — but the gap on agentic work is narrower than the version of this article written in September implies. For the flagship-against-flagship read, see Grok 4.7 vs GPT-5.6 Sol.

The API comparison below is incomplete, and it is incomplete in Grok's favour

This article argues the API advantage reversed, with GPT-5.6 Luna at $0.20 per million input against Grok at $2. That token comparison is correct. What it leaves out is that the thing this article credits Grok for — live X data — is not priced in tokens at all.

From xAI's own pricing page, search is billed as a tool invocation on top of token cost:

• X Search: $5 per 1,000 posts, or $10 per 1,000 profiles
• Web Search: $5 per 1,000 calls

So the workload this article recommends Grok for is the one workload where the token price is not the price. A sentiment sweep reading 2,000 posts costs $10 in search calls before a single token is counted.

Three more modifiers the original comparison omits:

The long-context tier is the one that catches people. It is not a surcharge on the tokens above 200k; once the prompt crosses that line, the whole request bills at the higher rate. Cached input helps in the other direction at $0.50 per million, or $1.00 on long context.

One note on the plan ladder

The tier table further down lists SuperGrok Lite at $10, SuperGrok at $30, SuperGrok Plus at $100 and SuperGrok Heavy at $300. Those were correct when written and xAI's own pages no longer agree with each other on them. The consumer pricing page currently shows only Free and SuperGrok at $30, naming Grok 4 — two generations stale — as the SuperGrok model, with Lite and Heavy appearing as unpriced columns in a comparison table. The X Subscription documentation shows a different ladder entirely: Plus at $8, Premium at $30, Super at $100 and Ultra at $200, with annual rates of $80, $300, $1,000 and $2,000.

We are not restating the tiers from a page that contradicts its own documentation. If you are choosing between the upper tiers, check the live page and read SuperGrok vs SuperGrok Heavy for what the jump actually buys. The one structural point that still holds, and which neither pricing page makes clear, is the shared usage pool: paid Grok plans draw Chat, Imagine, Voice and Build from one allowance, while ChatGPT meters each product separately.

Grok 4.7 figures from xAI's launch post and docs.x.ai pricing, both read 24 October 2026. The original 30-day test below is unchanged from the September version except where corrected above.

What Changed Since the Original Test

The original comparison ran in June. Four things have happened since, and two of them change the answer.

The Full Comparison

The API Argument Reversed

OpenAI split its line into three tiers, and the cheap one is genuinely cheap. For high-volume simple work — classification, extraction, routing — Luna is now the budget option and Grok is not.

One caveat that matters for budgeting: GPT-5.6 Sol currently sits at $4 and $20 promotionally, reverting to $5 and $30 around 21 November. Price any migration at the rate that will exist rather than the one that does.

The Shared Pool Nobody Mentions

Paid Grok plans run on a single weekly allowance covering Chat, Imagine, Voice and Build. Generate a run of images and you have less left for research. Spend an afternoon on voice and the same applies.

ChatGPT meters per product, so image generation does not consume your Codex budget. That difference does not appear on either pricing page and it is the thing most likely to surprise a heavy user in week two.

What 30 Days Revealed — Still True

The X advantage is real and narrower than you think

Asked what people were saying about NVDA earnings in the moment, Grok surfaced investor posts from the previous half hour. ChatGPT returned a news article from four hours earlier. For live sentiment on stocks, politics and breaking news, Grok wins and it is not close.

But that matters for a specific slice of work — journalists, social managers, political researchers, analysts tracking sentiment. Everyone else is paying $10 more for something they use twice a week.

ChatGPT wins on coding, and Astra widened it

For a developer using AI daily, ChatGPT Plus with Codex is the better $20 than SuperGrok alone at $30. Grok has no equivalent to Codex — it is a strong chat model that writes code, not an agent that runs a session.

Astra arriving on Plus makes that gap wider rather than narrower, and it costs subscribers nothing extra.

Grok voice is more natural, ChatGPT voice has more features

Grok at 300 to 500ms sounds more human, redirects cleanly when interrupted, and shifts tone appropriately. ChatGPT has more voices, desktop support and Siri handoff. For driving and casual use, Grok. For desktop workflows, ChatGPT. Both are good.

Four agent slots is still a real constraint

SuperGrok Custom Agents hold persona and workflow context better than ChatGPT custom GPTs, which drift. But four slots against unlimited is a genuine limit for anyone wanting agents across several functions. The workaround is here.

Pricing — Every Tier

Who Should Pay for What

FAQ

Related: ChatGPT Plus vs Pro now Astra ships · What GPT-6 Astra actually shipped with · SuperGrok Lite vs SuperGrok · Grok Custom Agents guide · Model release tracker