Grok vs Gemini: X-native agent or Google bundle?

Side by side

The table compares Grok from SpaceXSI (formerly xAI) with Google’s Gemini. Prices are US monthly prices before tax. Leaderboard positions are as of October 15-18, 2026.

Ranks in the last two rows are as of October 15-18, 2026. A token is a small chunk of text that AI pricing is measured in.

Where Grok tends to win

Lower output price on the API

Both flagships cost $2 per million input tokens today. On output, Grok 4.7 costs $6 per million tokens against $10 for Argon at its introductory rate. Google says Argon will rise to $4 / $20. Artificial Analysis notes that Grok 4.7 is very verbose, so real costs per task can be higher than the list price suggests.

An always-on agent from $30

Grok Bot runs on a cloud computer that keeps working while your laptop is off, and it comes with the $30 Premium plan. Google’s closest offers, Agent Mode and Gemini Spark, go to AI Ultra subscribers first, and Ultra starts at $99.99.

X built in

Grok lives inside X, and its paid plans include X Premium perks. Grok Bot can read and act on your X messages once the account is connected. If X is where you spend your time, that matters more than any benchmark.

Beats Gemini’s fast model

On the Artificial Analysis index, Grok 4.7 scores 46 against 41 for Gemini 3.8 Flash. Flash is what Gemini’s cheaper plans run on, since AI Plus lost Pro-model access in October.

Where Gemini tends to win

Higher independent scores

Gemini 4 Argon scores 53 on the Artificial Analysis Intelligence Index, 7 points above Grok 4.7. It is #1 on the Arena Text board, though still marked preliminary, while Grok 4.7 sits at #89. Artificial Analysis also says Argon matches GPT-6 Astra on its index at about 60% of the cost per task when discounted.

Cheaper entry and the Google bundle

AI Plus costs $4.99, and AI Pro at $19.99 includes Argon and 5 TB of Google storage. Plans also bundle Gemini in Gmail and Docs, NotebookLM and YouTube Premium (Lite on Pro, full on Ultra). If you already pay for Google storage, Gemini can be close to free.

One open question: Google’s October 13 post names only Pro and Ultra as getting Argon. A footnote on Google’s site suggests Plus might too. We could not confirm it.

Longer context and more input types

Argon takes about 1.05 million tokens of context, roughly double Grok 4.7. It accepts text, images, video, audio and PDF as input.

Images and video: close to a tie

Both make images and video. Gemini uses Nano Banana for images, even on the free tier, and Omni for clips up to 10 seconds on paid plans. Grok uses Grok Imagine, with video from Premium up. Both have had trouble: Google pulled Nano Banana from Google Earth within 24 hours in July 2026 over misinformation concerns, and Grok’s image tools drew regulator investigations over sexual deepfakes.

An honest note on benchmarks

• Arena scores differ between views. One outlet reported Argon at 1532 on the October 15 board; the live page showed 1524. The rank, #1, was the same.
• Artificial Analysis ranks depend on how effort settings are counted, which is why Grok 4.7 appears anywhere from about 25th to 30th.
• Google’s own benchmark claims for Argon, such as 77.9% on DeepSWE, are the maker’s numbers.
• Introductory API prices are scheduled to rise, so check the current rate before you build.

Which should you pick?

• Pick Grok if you use X a lot, want an always-on agent without paying for a top tier, or want the lower output price on the API.
• Pick Gemini if you live in Gmail, Drive and Docs, want the storage bundle, need a long context window, or want the model at the top of the current rankings.
• Mind the free tiers. Gemini Free now runs only Flash-Lite. See what Grok’s free plan includes.

We also compare Grok with ChatGPT and Claude.

Questions people ask