Blog

Grok vs Gemini in 2026: models, plans, benchmarks and API prices compared

Ramashankar S · · Updated · 17 min read

For most people, Gemini is the better pick in the Grok vs Gemini choice: Google AI Pro costs $19.99 a month and includes Gemini 4 Argon, which ranks second on the Arena text leaderboard and scores 50 on the Artificial Analysis Intelligence Index, against 35 for Grok 4.7. Grok is the better pick if you live on X, want Cursor and X Premium in the same $30 bill, or build on the API, where Grok 4.7 costs $2 / $6 per million tokens against Argon's $4 / $20 from 1 December 2026.

Below, the full Grok vs Gemini comparison for 2026: the models each app gives you, every plan, how usage limits work, independent benchmarks, privacy, and API prices for builders. Prices and limits are from xAI's and Google's own pages, checked on 10 November 2026.

Grok vs Gemini: the short answer

If you wantPickWhy
The strongest model for the moneyGemini (AI Pro, $19.99)Argon leads Grok 4.7 on Arena and Artificial Analysis
A free AI chat appGemini, narrowlyFree Gemini has Deep Research and Gemini Live, but only Flash-Lite
Real-time posts from XGrokGrok searches X directly, and the plan includes X Premium
Gmail, Docs and Drive helpGeminiAI Pro adds Gemini to Gmail and Docs plus 5 TB of storage
One bill for chat and a coding editorGrok (Premium, $30)Includes Cursor Pro and Grok Bot from one usage pool
The lowest API price on a frontier modelGrok 4.7$2 / $6 against Argon's $4 / $20 from 1 Dec
The lowest API price overallGemini 3.8 Flash$0.75 / $3.75 until 31 Dec, then $1.50 / $7.50
Grok vs Gemini plans side by side: The X Subscription from $8 to $200 a month and Google AI from $4.99 to $199.99

Which models you get in each app

Gemini's models depend on your plan. Grok's app doesn't publish a model list per plan; xAI's pricing page says only that Premium and above get the "Latest Grok models".

Grok (xAI)Gemini (Google)
Top public modelGrok 4.7, released 21 Sep 2026Gemini 4 Argon, in the app since 13 Oct 2026
Fast modelNot named per planGemini 3.8 Flash
Free planGrok, with limited chat requestsFlash-Lite only (since 9 Oct)
Context window500,000 tokens on the API1M tokens in AI Pro (app)
Newer, restricted modelNone announcedGemini 4 Carbon, cyber defenders only

Google's Argon announcement puts the model in the Gemini app for AI Pro and AI Ultra. 9to5Google reported that the AI Pro model picker shows 3.5 Flash-Lite, 3.8 Flash, 3.1 Pro and 4 Argon. Free accounts lost Flash and Pro on 9 October and now get Flash-Lite only, and AI Plus subscribers lose manual access to Pro models in November (TechRadar, Android Headlines).

Grok 4.7 is xAI's newest model, built on a larger base model than Grok 4.6 and aimed at coding and knowledge work. xAI (which now signs its pages SpaceXSI) also serves a Fast variant at twice the speed and twice the price, but only inside Cursor and Grok Build. Gemini 4 Carbon, announced 28 October, outscores everything else from either company, but it isn't in the Gemini app or the API yet.

Grok vs Gemini plans and prices

Both companies now sell four paid tiers. Google's start lower; xAI's include more than chat.

Grok: The X SubscriptionMonthlyAnnualGemini: Google AIMonthly
Free$0Free$0
Plus$8$80AI Plus$4.99
Premium$30$300AI Pro$19.99
Super$100$1,000AI Ultra (5x Pro)$99.99
Ultra$200$2,000AI Ultra (20x Pro)$199.99

US prices before tax. Sources: xAI pricing and X Subscription plan docs; Google AI plans.

What each tier adds:

  • Grok Plus ($8) gives Grok chat, voice, image generation and Grok Build, Cursor access, Grok API access, and X's checkmark with fewer ads. It has no Grok Bot.
  • Grok Premium ($30) has 4 times Plus usage, Grok Bot, the latest Grok models, image and video generation, Cursor's cloud agents and code reviews, and ad-free X on most surfaces. xAI says it bundles what SuperGrok, Cursor Pro and X Premium+ cost together at $90.
  • Grok Super ($100) and Ultra ($200) raise usage to 15 and 50 times Plus. Super adds 1080p, 15-second videos and priority at peak times.
  • Google AI Plus ($4.99) gives the Flash model, twice the free usage, video generation, 400 GB of storage and Gemini in Gmail and Vids.
  • Google AI Pro ($19.99) adds Gemini 4 Argon and the Pro model, 4 times free usage, a 1M-token context window, 5 TB of storage, Gemini in Gmail and Docs, Chrome auto browse and YouTube Premium Lite.
  • Google AI Ultra ($99.99 or $199.99) gives 5 or 20 times Pro usage, the most Argon access, 20 TB of storage, Tasks, Project Genie and Veo 3.1 video.

The X Subscription launched on 12 October 2026, and grok.com's plans page now shows only its tiers. Existing SuperGrok, Cursor and X Premium subscribers can keep their old plan or switch for a prorated refund.

At the entry tier the two compare cleanly: $19.99 gets Argon, the model that leads the leaderboards below, while $30 gets Grok 4.7 plus Cursor Pro and X Premium+. Cursor's own pricing page now sells the same five tiers, so if you'd pay for Cursor anyway, Grok and X Premium come with it. If you only want a chatbot, AI Pro costs $10 less and has the stronger model.

Usage limits and running out

Neither company publishes message counts. Both measure compute and reset on a schedule:

  • Gemini's limits "factor in the complexity of your prompt, the features you use, and the length of your chat". They refresh every 5 hours until you reach a weekly cap, and you can buy more usage (Google AI plans).
  • Grok's paid plans get one weekly pool shared across Grok chat, Imagine, voice, Grok Bot and Cursor. Past the pool, further use is paid from a prepaid balance, with optional spend caps (X Subscription usage docs).

The shared pool cuts both ways. A long Cursor session draws down the same allowance you chat with, and Claude or GPT models in Cursor are never covered: they're billed at API list prices from the first request. xAI's announcement says each X Subscription tier includes at least twice the coding usage with Grok models of the comparable older Cursor plan.

Grok vs Gemini benchmarks

Any Gemini vs Grok AI comparison in 2026 has to separate independent tests from vendor charts. On independent leaderboards Gemini leads by a wide margin. Grok 4.7 does better on xAI's own coding and professional-work tests.

BenchmarkGrok 4.7Gemini 4 ArgonGemini 3.8 Flash
Arena text, rank and score (9 Nov)#94, 1442#2, 1522#9, 1497
Artificial Analysis Intelligence Index35 (high and xhigh)50 (high)31 (high)
AA cost per index task$6.18 (high)$9.95 (at $4 / $20)$7.77
AA time per index task21.7 min (high)29.2 min14.8 min
Vals Index#24, 54.2%#1 (Google's claim)
CursorBench 4.0 (xAI's figure)46.3%
Terminal-Bench 4.0 (xAI's figure)37.6%

Sources: Arena text leaderboard, Artificial Analysis leaderboard, Grok 4.7 announcement, Argon announcement.

Grok 4.7 vs Gemini 4 Argon and 3.8 Flash on the Arena text leaderboard and the Artificial Analysis Intelligence Index

Three caveats. The best-placed Grok on Arena is an older one: Grok 4.20 Beta 1 at #38 (1475), well above Grok 4.7's #94. Artificial Analysis has rescaled its index since launch, when The Decoder reported Grok 4.7 at 46 on the older version; on today's scale Argon still leads by 15 points. And cost per task depends on how much a model thinks: Grok 4.7 is cheap per token but writes long reasoning, so it costs $6.18 to $7.77 per index task, close to Argon at the higher December price.

Gemini 3.8 Flash ranks ninth on Arena, 85 places above Grok 4.7, so even the Flash model on the $4.99 AI Plus plan rates above Grok's newest model with Arena voters.

Gemini 4 Argon: What's new and what's nextGoogle for Developers · 30:21 · YouTube
Google's launch session on Gemini 4 Argon, including where it's available and its 1M-token output limit.

Features: search, X, video and your Google account

FeatureGrokGemini
Live web searchYesYes, Google Search
Live X postsYes, nativeNo
Deep research reportsNot listed on plan pagesDeep Research, free plan included
Voice modeGrok VoiceGemini Live, free
Image generationGrok Imagine, limited on freeYes, with Nano Banana Pro limited on free
Video generationPremium and up; 1080p 15 s on SuperAI Plus and up; Veo 3.1 on Ultra
Email and documentsNoGmail and Docs on AI Pro
Coding editorCursor, from PlusAntigravity (Argon needs Ultra)
StorageNone400 GB to 20 TB+
Social perksX checkmark, fewer or no adsYouTube Premium Lite on AI Pro

Grok's one feature Gemini can't match is direct access to X. The xAI API sells the same thing as an x_search tool, which shows how central it is. Gemini's advantages are breadth: it lives inside Gmail, Docs, Chrome and Android, and a paid plan doubles as Google One storage.

For coding, Gemini's 9to5Google report says Argon in the Antigravity desktop app and CLI requires AI Ultra. Grok Premium includes Cursor with Grok 4.7, and Cursor can run Claude and GPT models at API prices. The Cursor vs Claude Code post covers that editor in detail.

Privacy in the apps

Both use consumer chats to improve their products by default:

  • Google uses Gemini chats for improving its services, including training generative models, while Keep Activity is on. A subset is read by human reviewers and kept up to three years. With Keep Activity off, chats are kept 72 hours (Gemini Apps Privacy Hub).
  • xAI's privacy policy (updated 27 October 2026) lists developing and improving its service among the uses of your prompts and outputs. Private Chat conversations stay out of your history and are deleted within 30 days. Controls are under Settings, Data Controls.

API traffic is different: both companies say paid API data isn't used for training, as covered in the next section.

Grok vs Gemini API prices for builders

On the API, Grok 4.7 costs less than Gemini 4 Argon per token, and Gemini 3.8 Flash costs less than either until 31 December. Prices per million tokens, standard tier:

ModelInput / outputCached inputLong promptsBatchContext
grok-4.7$2.00 / $6.00$0.50$4.00 / $12.00 from 200K tokens20% off500K
grok-4.3$1.25 / $2.50$0.20$2.50 / $5.00 from 200K20% off1M
gemini-4-argon, to 30 Nov$2.00 / $10.00$0.10Same rate50% off1M
gemini-4-argon, from 1 Dec$4.00 / $20.00$0.20Same rate50% off1M
gemini-3.8-flash, to 31 Dec$0.75 / $3.75$0.075Same rate50% off1M
gemini-3.8-flash, from 1 Jan 2027$1.50 / $7.50$0.15Same rate50% off1M

Sources: xAI models, xAI pricing, Gemini API pricing.

The details that change a bill:

  • Grok bills the whole request at the higher rate once the prompt reaches 200,000 tokens. Argon and 3.8 Flash charge one rate across their 1M window.
  • Cache hits cost 75% less on Grok and 95% less on Gemini, so cache-heavy agents narrow Grok's price lead.
  • Web search costs $5 per 1,000 calls on xAI. Google gives 5,000 grounded searches a month free across Gemini 3 and newer, then charges $14 per 1,000, with a cap of 1,500 a day per project.
  • Reading X posts through x_search costs $5 per 1,000 posts returned, counting quoted and parent posts.
  • Gemini's free API tier covers 3.8 Flash and the Flash-Lite models but not Argon. xAI's rate limits page mentions no free allowance; its Tier 0 still allows 150 requests a second on grok-4.7.
  • Google's paid tiers cap monthly spend at $250 until you've paid $100 and waited 3 days, then $2,000 until $1,000 and 30 days (Gemini billing). Its prepaid balance expires after a year.

Two monthly bills at list prices, from the script in this post's notes (prompts under 200,000 tokens; output includes thinking tokens, and real counts differ by model):

Monthly API cost for a research assistant and a document Q&A workload on Grok 4.7, Grok 4.3, Gemini 4 Argon and Gemini 3.8 Flash

Research assistant: 40,000 requests a month, 4,000 input tokens each with no caching, 1,000 output tokens and one web search per request.

ModelTokensSearchMonth
grok-4.3$300.00$200.00$500.00
grok-4.7$560.00$200.00$760.00
gemini-3.8-flash, Dec$270.00$490.00$760.00
gemini-3.8-flash, from Jan$540.00$490.00$1,030.00
gemini-4-argon, Nov$720.00$490.00$1,210.00
gemini-4-argon, from Dec$1,440.00$490.00$1,930.00

Document and code Q&A: 20,000 requests, 30,000 input tokens each with 80% cache hits, 2,000 output tokens, no search.

ModelMonth
gemini-3.8-flash, Dec$276.00
grok-4.3$346.00
gemini-3.8-flash, from Jan$552.00
gemini-4-argon, Nov$688.00
grok-4.7$720.00
gemini-4-argon, from Dec$1,376.00

From December, Grok 4.7 costs about half what Argon does on both jobs. Search-heavy apps favor Grok even against 3.8 Flash, because Google's grounding fee is almost three times xAI's. For cached document work, 3.8 Flash costs least through December. Run the numbers for your own traffic with the token cost calculator, and see the Gemini API pricing post for Google's full price list and the best LLM API comparison for rate limits, billing and data terms across six providers.

On data: xAI stores API requests and responses for 30 days and never trains on them without permission; zero retention is a team-wide setting (xAI security FAQ). Google doesn't train on paid-tier API data but uses free-tier prompts to improve its products.

Grok and Gemini through Aight

Aight is one OpenAI-compatible API (https://api.getaight.ai/v1) for 138 models, including grok-4.7 at xAI's $2 / $6 and gemini-3.1-pro-preview at Google's $2 / $12. One key lets you send the same prompts to both makers and compare answers and bills before you commit.

Where going direct works better:

  • Aight's models page charges grok-4.7's higher price from 128,000 tokens, against 200,000 at xAI, so prompts between the two cost less direct.
  • Aight lists Gemini 4 Argon at $4 / $20 and 3.8 Flash at $1.50 / $7.50, the prices Google moves to on 1 December and 1 January, so both cost less on Google's API until then.
  • Aight documents chat completions, embeddings and the model list only, so xAI's x_search, Google's search grounding and both Batch discounts are direct-only.

Aight's balance never expires and unspent paid balance is refundable. A $100 recharge costs $103 outside India and adds $110 of balance. The one API key post shows the setup.

Comparing either with OpenAI instead? See Gemini vs ChatGPT and Grok vs ChatGPT. And if you meant Groq, the company that serves open models on its own chips, Groq vs Grok explains the difference.

What developers report

  • "Significant regression in token efficiency compared to Grok 4.6" (Hacker News commenter on the Grok 4.7 launch, 21 Sep 2026, link)
  • "The token allowance with a Cursor subscription is crazy." (Hacker News commenter on Grok 4.7, 21 Sep 2026, link)
  • A Hacker News commenter said Grok's price for cache hits was double other providers', a problem for cache-heavy work (21 Sep 2026, link)
  • "It seems the model is a little verbose" (Hacker News commenter on Gemini 4 Argon, citing Artificial Analysis, 13 Oct 2026, link)
  • A Hacker News commenter noted that Argon launched at GPT-6 Sol's price and doubles on 1 December to Claude Opus 5.5's (13 Oct 2026, link)
  • A Hacker News commenter complained that the free Gemini app dropped Flash and left only Flash-Lite (21 Oct 2026, link)

FAQ

Is Grok better than Gemini?

Not on independent benchmarks. Gemini 4 Argon ranks second on the Arena text leaderboard and scores 50 on the Artificial Analysis Intelligence Index; Grok 4.7 ranks 94th and scores 35. Grok wins on API price, live X data and its bundle with Cursor and X Premium.

Is Grok free?

Yes, with limits. The X Subscription's free plan includes daily access to Grok, a starter allowance for Cursor's agents and standard X. xAI doesn't publish the limits or say which model free users get.

Is Gemini free?

Yes. The free Gemini app includes Deep Research, Gemini Live, Canvas and Gems, but only the Flash-Lite model since 9 October 2026. The Gemini API also has a free tier for 3.8 Flash and the Flash-Lite models, and Google may use those prompts to improve its products.

Which costs less, SuperGrok or Gemini Pro?

Gemini. Google AI Pro is $19.99 a month. SuperGrok's successor for new customers is The X Subscription Premium at $30 a month, or $300 a year, which also includes Cursor Pro and X Premium+.

Which is better for coding, Grok or Gemini?

Argon scores higher on independent tests, but in Antigravity it needs AI Ultra at $99.99 or more. Grok Premium at $30 includes Cursor with Grok 4.7 and a weekly pool xAI says holds at least twice the coding usage of the older comparable Cursor plan. The best LLM for coding post ranks both against Claude and GPT.

Does Grok or Gemini have a bigger context window?

Gemini. Argon, 3.8 Flash and AI Pro in the app all handle 1M tokens. Grok 4.7 takes 500,000 tokens on the API, and Grok 4.3 takes 1M.


Sources

Checked on 10 November 2026. Updated 17 November 2026: Aight's current model list.