Grok (xAI) Statistics 2026 — 600M Grok Users Is the Wrong Headline
Grok has gone from a feature inside X to a standalone AI platform, enterprise product, developer API, voice stack and multimodal creation suite. But the headline numbers need careful handling: xAI’s January 2026 disclosure of approximately 600 million monthly active users covers X and Grok together, not 600 million Grok-only users.
This report compiles the most decision-useful Grok user statistics, xAI valuation and funding data, Grok AI usage statistics, model benchmarks, infrastructure metrics and global deployment figures available through August 26, 2026.
Monthly active users across the combined X and Grok apps — not a Grok-only MAU figure.
Grok.com visits in December 2025, up from 1.2M in January.
xAI’s upsized Series E financing completed in January 2026.
Reported valuation assigned to xAI in SpaceX’s February 2026 acquisition.
GPU scale of Colossus; a May 2026 update put Colossus 1 above 220,000 NVIDIA GPUs.
Grok 4.6 score on the Artificial Analysis Intelligence Index — tied with GPT-5.6 Sol.
// Table of Contents
The Most-Repeated Grok Statistic Needs a Denominator
The biggest Grok number in circulation is also the easiest to misuse. On January 6, 2026, xAI said its reach spanned approximately 600 million monthly active users across the X and Grok apps. That is a distribution metric for the combined ecosystem, not a standalone Grok MAU disclosure.
That distinction matters for founders and marketers benchmarking AI platforms. X supplies an enormous built-in surface for Grok, while grok.com and the mobile app are separate destinations with their own engagement. Treating all three as one “Grok users” figure inflates product-level adoption and makes comparisons with ChatGPT, Claude or Perplexity unreliable.
Monthly active users across X and Grok apps in xAI’s January 2026 company update.
Estimated worldwide Android downloads for Grok in Sensor Tower’s latest month.
Estimated worldwide Android revenue for Grok in Sensor Tower’s latest month.
Google Play install range shown for Grok’s Android app in August 2026.
Grok’s U.S. Google Play Productivity rank on the Top Free chart in the cited Sensor Tower snapshot.
Grok’s U.S. Google Play Productivity rank on the Top Grossing chart in the same snapshot.
Do not publish “Grok has 600M monthly users” as a direct product statistic. xAI’s wording is “approximately 600 million monthly active users across the X and Grok apps.” The defensible wording preserves that combined scope.
The mobile app still shows real standalone scale. Sensor Tower’s August 2026 page places Grok in the 100–500 million Google Play install range and estimates four million worldwide Android downloads plus $10 million in Android revenue in the latest month. Those are estimates, not xAI audited figures, but unlike the 600 million headline they measure a Grok-specific surface.
Range of in-app purchase prices shown on Grok’s U.S. Android listing.
Worldwide Android release date listed for the Grok app.
“Starlink was already a cash flow engine and now it adds an AI revenue layer on top.”Ali Javaheri, PitchBook Senior Emerging Spaces Analyst, quoted by Reuters · Feb. 2026
That quote captures the broader strategic shift. Grok is no longer just an assistant competing for chat sessions; after SpaceX acquired xAI, the AI business sits inside a larger infrastructure-and-distribution system spanning X, Starlink, data centers, government deployments and developer APIs.
Grok.com Grew 22,388% in 2025 — But Not in a Straight Line
Similarweb’s full 2025 series is the cleanest public trend dataset for Grok’s standalone web destination. It shows explosive launch-driven acquisition, a mid-year retreat, and a stronger Q4 finish.
The shape is more useful than the endpoint. Grok.com jumped from roughly 1.2 million January visits to 51.5 million in February and 202.7 million in March. It then fell for three consecutive months, recovered in July, softened again through September, and finally closed December at 271.2 million visits.
Grok.com’s February 2025 month-over-month traffic growth, according to Similarweb.
Additional Grok.com traffic growth in March 2025 after February’s surge.
Total Grok.com growth from January to December 2025.
Grok’s traffic did not simply collapse after launch. December 2025 set the year’s high at 271.2M visits, about 34% above the March breakout level. The mid-year decline was a retention warning, but Q4 showed the standalone web product could rebuild momentum.
For operators, this is a reminder to separate launch velocity from habit formation. Grok’s first two growth months were extraordinary; the strategic question was whether product breadth, model quality and distribution could turn that attention into repeat usage. The rest of 2026’s product roadmap looks designed around exactly that problem.
Why Grok Grew: Distribution, Compute, Product Breadth and Price
No single factor explains Grok’s rise. The data points to four compounding advantages: X distribution, unusually fast compute buildout, frequent model releases, and aggressive API economics.
Combined monthly active reach across X and Grok gave xAI an acquisition surface most standalone AI startups cannot reproduce.
Grok.com traffic jumped more than fortyfold in the month Grok 3 launched.
xAI says Colossus reached its first 100K-GPU operating scale in 122 days.
Colossus was then doubled to roughly 200K GPUs in another 92 days.
Grok 4.6 starts at $2 per million input tokens, with output at $6 per million.
Grok Imagine was ranked first on the cited Artificial Analysis text-to-video table in xAI’s 2026 product page.
The first driver is distribution. xAI’s acquisition of X in March 2025 valued the social platform at $33 billion after debt and put the AI company and the network under one roof. The second is infrastructure: xAI repeatedly compressed data-center timelines that normally take years into months.
Value assigned to X after debt in xAI’s March 2025 all-stock acquisition.
Value assigned to xAI in the same March 2025 transaction.
NVIDIA Hopper GPUs in Colossus when xAI announced its Series C in December 2024.
Time from first server delivery until Colossus began running workloads, per xAI.
Compute-efficiency improvement xAI reported for Grok 4 training infrastructure/algorithms.
Increase in training compute for Grok 4 versus prior runs, according to xAI.
xAI did not grow by choosing between product distribution and compute. It funded both. The sequence runs from a $6B Series B, to another $6B Series C, to $10B of debt + equity in mid-2025, to a $20B Series E in January 2026.
The fourth driver is product breadth. By mid-2026 Grok was no longer a text chatbot: xAI was selling reasoning, coding, web/X search, image and video generation, speech-to-speech voice, text-to-speech, speech-to-text, enterprise security and agent tooling. Each surface creates another route to usage—and another reason the company-level reach number cannot be reduced to a single chatbot MAU.
For growth teams, that breadth changes how competitive share should be measured. A user who invokes Grok inside X, calls the API from an agent, generates a video in Imagine and later uses the standalone app may touch four products while still being one person. The useful operating view is therefore surface-by-surface: acquisition by channel, repeat usage by workflow, paid conversion by product and compute cost per successful outcome. That framework is less headline-friendly than one giant user number, but it is far more actionable.
Consumer, Developer, Enterprise and Government Grok Are Different Businesses
A single “Grok adoption” number hides four distinct demand pools. Consumer usage is measured in visits, installs and subscriptions; developer usage is priced by token and tool call; enterprise is sold by seat and security tier; government is distributed through institutional programs.
Self-serve Grok Business price for small and medium teams at launch.
Context window for Grok 4 Fast reasoning and non-reasoning API variants.
Grok 4.1 Fast input-token price per million tokens.
Cached-input price per million tokens for Grok 4.1 Fast.
Grok 4.1 Fast output-token price per million tokens.
Starting price for 1,000 successful tool invocations with Agent Tools API.
Audio pricing for Grok Voice Think Fast 2.0.
U.S. military and civilian DOW employees targeted for GenAI.mil access to xAI technology.
Impact Level 5 environment cited for the Department of War deployment.
Enterprise packaging is intentionally separate from raw API consumption. Grok Business launched at $30 per seat per month, while Enterprise added custom SSO, SCIM directory sync and advanced controls. xAI also says enterprise customer data is not used for training, and its Vault tier adds customer-managed encryption keys and a dedicated data plane.
Web visits, mobile installs, API tokens, paid seats and government access are not additive user counts. They overlap heavily. A decision-quality dashboard should track each as its own funnel rather than summing them into a fictional “total Grok users” number.
The developer segment has moved from subsidized experimentation to aggressive price/performance competition. xAI’s 2024 API beta offered $25 in free monthly credits and a 128,000-token context window. By late 2025, Grok 4 Fast expanded context to two million tokens while pricing sub-128K input at $0.20 per million tokens and output at $0.50 per million.
Free API credits offered to every participant during xAI’s 2024 public beta.
Context length of the grok-beta model in the November 2024 API launch.
Grok 4 Fast input price at context lengths of 128K tokens or more.
Grok 4 Fast output price at context lengths of 128K tokens or more.
xAI’s Competitive Strategy Is Capital-Intensive by Design
The company’s funding history shows a deliberate choice to compete on infrastructure scale. That strategy culminated in a $20 billion Series E and then a reported $250 billion valuation in SpaceX’s February acquisition.
Debt financing completed in mid-2025, according to Morgan Stanley via Reuters.
Strategic equity financing completed alongside the 2025 debt raise.
Original target size of the Series E before it was upsized to $20B.
Reported xAI value in the February 2026 SpaceX acquisition.
Reported SpaceX value used in the same acquisition transaction.
SpaceX shares offered per xAI share under the reported merger terms.
Cash-per-share option reported for certain xAI executives.
xAI’s last reported November 2025 valuation cited by Reuters.
SpaceX’s last private valuation before the deal, cited by Reuters.
The valuation jump was extremely fast. Reuters reported xAI at $80 billion in the March 2025 transaction that absorbed X, around $200 billion after a September 2025 financing report, $230 billion in November, and $250 billion in the February 2026 SpaceX acquisition. Those figures come from different transactions and reporting dates, so they should not be presented as a single audited valuation series.
Post-money valuation reported by CNBC and carried by Reuters in September 2025.
H100 GPU equivalents across Colossus I and II by the end of 2025, according to xAI.
NVIDIA GPUs in Colossus 1 by May 2026, including H100, H200 and GB200 accelerators.
Capital intensity cuts both ways. In May 2026, SpaceXAI agreed to provide Anthropic access to Colossus 1. Reuters Breakingviews reported the arrangement involved 300MW of unused capacity—evidence that owning extraordinary compute does not automatically guarantee equally extraordinary utilization.
For investors and operators, the useful metric is therefore not “GPU count” alone. It is GPU utilization, inference demand, model monetization and the cost of turning infrastructure into durable revenue. The SpaceX acquisition makes that question larger, because AI compute can now be monetized through internal Grok workloads, third-party customers and potentially future orbital infrastructure.
Grok 4.6 Is an Agentic Model Story, Not Just a Chatbot Upgrade
By August 2026, xAI’s product frontier had shifted toward long-running agents, coding, tool use, voice and multimodal generation. Grok 4.6 is the clearest expression of that change.
| Benchmark | Grok 4.6 | Grok 4.5 | GPT-5.6 Sol | Fable 5 | Source |
|---|---|---|---|---|---|
| AA Intelligence Index | 61 | 56 | 61 | 62 | SpaceXAI · Introducing Grok 4.6 · Aug. 12, 2026 |
| GDPVal-AA v2 | 1753 | 1526 | 1728 | 1741 | SpaceXAI · Introducing Grok 4.6 · Aug. 12, 2026 |
| CursorBench v3.2 | 69.9% | 66.7% | 67.2% | 70.5% | SpaceXAI · Introducing Grok 4.6 · Aug. 12, 2026 |
| DeepSWE v1.1 | 65.9% | 54% | 73% | 70% | SpaceXAI · Introducing Grok 4.6 · Aug. 12, 2026 |
| FrontierCode v1.1 | 61.3% | 56.6% | 60.6% | 63.6% | SpaceXAI · Introducing Grok 4.6 · Aug. 12, 2026 |
| APEX-Agents | 57.5% | 47.1% | 56.7% | 59.2% | SpaceXAI · Introducing Grok 4.6 · Aug. 12, 2026 |
| Terminal-Bench v3.0 | 26% | 15.7% | 34.6% | 34.1% | SpaceXAI · Introducing Grok 4.6 · Aug. 12, 2026 |
| APEX-SWE | 56.4% | 53.6% | — | 58.8% | SpaceXAI · Introducing Grok 4.6 · Aug. 12, 2026 |
| AA-Briefcase | 1577 | 1313 | 1502 | 1574 | SpaceXAI · Introducing Grok 4.6 · Aug. 12, 2026 |
| Harvey LAB (Vals) | 15.8% | 12.9% | 2.5% | 11.3% | SpaceXAI · Introducing Grok 4.6 · Aug. 12, 2026 |
Grok 4.6 tied GPT-5.6 Sol at 61 on the Artificial Analysis Intelligence Index and exceeded it on GDPVal-AA v2, CursorBench, FrontierCode, APEX-Agents, AA-Briefcase and Harvey LAB in xAI’s published comparison. It trailed on DeepSWE and Terminal-Bench. That mixed picture is more informative than any single “best model” claim.
Grok 4.6 starting input-token price per million tokens.
Grok 4.6 starting output-token price per million tokens.
Fast Grok 4.6 variant is priced at twice the standard variant.
Included usage promotion in Grok Build and Cursor for the first week of launch.
Average output tokens Grok 4.5 used per SWE Bench Pro task.
Average output tokens for Opus 4.8 Max on the same xAI comparison.
Output-token advantage xAI reported for Grok 4.5 vs. Opus 4.8 Max on SWE Bench Pro.
Grok’s differentiation is moving from “personality + X access” toward integrated agent economics: long context, tool use, web/X search, coding, voice, image/video generation and lower token consumption per completed task.
The adjacent technology stack is equally important. Grok Voice Think Fast 2.0 reached 82.9% on Artificial Analysis’s speech-to-speech quality index, while Grok Imagine won 57% of xAI’s side-by-side overall editing comparison against Kling o1 and 64.1% against Runway Aleph. These are vendor-published evaluations and should be read with that caveat, but they show where xAI is investing.
Grok Voice Think Fast 2.0 on the AA Speech-to-Speech Quality Index.
Speech Reasoning Big Bench Audio score for Grok Voice Think Fast 2.0.
Full Duplex conversational-dynamics score for Grok Voice Think Fast 2.0.
τ-voice agentic performance score for Grok Voice Think Fast 2.0.
Time to first audio for Grok Voice Think Fast 2.0 in the published comparison.
Relative median reasoning tokens per response vs. 1.0× for Voice Think Fast 1.0.
Languages used in xAI’s transcription-accuracy evaluation.
Claimed transcription advantage in noisy settings versus dedicated STT models.
Grok Imagine overall preference vs. Kling o1 in xAI’s IVEBench editing comparison.
Grok Imagine overall preference vs. Runway Aleph in the same evaluation.
Maximum images per request listed for the Imagine image API.
Supported image output resolution tiers listed on the Imagine API page.
The Human Layer: Users, Developers, Employees and Institutional Buyers
Grok’s most consequential “people statistics” are not demographics; xAI does not publish a robust age/gender breakdown. The stronger data is about who can access the system, how developers consume it and which institutions are operationalizing it.
Potential military and civilian DOW users under the GenAI.mil deployment.
Students targeted by the two-year El Salvador education deployment.
Public schools included in the El Salvador rollout.
Planned deployment period for the nationwide education program.
Telegram’s user base cited when xAI agreed to distribute Grok through the messaging app.
Cash-and-stock consideration xAI agreed to pay Telegram under the one-year partnership.
Telegram’s share of Grok subscription sales generated through the app under the reported agreement.
Length of the original Telegram–xAI distribution agreement.
For developers, Grok’s economics are increasingly workload-specific. Grok 4.1 Fast’s research benchmark is a good example: xAI reported a 63.9 score on Reka Research Eval at an average cost of $0.046, 87.6 on FRAMES at $0.048, and 56.3 on its internal X Browse benchmark at $0.091.
Grok 4.1 Fast Agent Tools score on Reka Research Eval.
Average cost reported for that Reka Research Eval run.
Grok 4.1 Fast score on xAI’s internal X Browse benchmark.
Average cost reported for that X Browse run.
xAI says Grok 4.1 Fast cut hallucination rate roughly in half versus Grok 4 Fast.
Token price is not enough. For agentic workloads, cost per completed research task, tool-call reliability, latency and output-token efficiency can dominate the nominal $/million-token rate.
Grok 4.5 makes the same point from another angle. xAI reported 80 tokens per second and roughly 4.2× fewer output tokens than Opus 4.8 Max on SWE Bench Pro. Even when two models post similar benchmark accuracy, the model requiring fewer steps and tokens can produce a materially different unit-economics outcome.
Grok’s Expansion Is Becoming Institutional, Not Just App-Store Global
The geographic story in 2026 is unusually diverse: consumer app demand is strongest in large mobile markets, while xAI is simultaneously pursuing national-government deployments, military access and data-center partnerships.
India
Top country/region listed for Grok’s Android app in Sensor Tower’s August snapshot.
United States
Second of the three top countries/regions displayed in the same Sensor Tower snapshot.
Indonesia
Third top country/region displayed for the Grok Android app.
Saudi Arabia
xAI announced nationwide Grok deployment across public and private entities through its HUMAIN framework.
El Salvador
Grok education program targets 5,000+ public schools and 1M+ students over two years.
U.S. Government
GenAI.mil deployment is designed to make xAI tools available to 3M DOW personnel at IL5.
Saudi Arabia is the clearest example of a country-level strategy rather than a conventional software sale. xAI’s November 2025 framework with the Kingdom and HUMAIN calls for a national AI layer, nationwide Grok deployment and new hyperscale GPU data centers. No public user count was disclosed, so this article does not turn the phrase “nationwide” into an invented adoption number.
Palo Alto, Seattle, Memphis and London are listed as SpaceXAI/xAI office locations on the current company page.
Voice workflows are described as operating across more than 25 languages.
Combined X + Grok MAU provides the distribution backdrop for global rollouts, but remains a combined metric.
Country deployment and country usage are not the same. Saudi Arabia’s partnership is national in scope, while Sensor Tower’s country labels reflect where the Android app is strongest. Neither is a substitute for audited country-level MAU.
The Data Says Distribution Alone Is Not Enough
Grok has the rare combination of a built-in social network, massive compute and deep financing. Yet the benchmark and traffic data show why execution still matters: launch spikes cool, frontier rankings move quickly, and infrastructure can become underutilized.
Grok 4 Fast’s search profile demonstrates the value of native retrieval. xAI reported 95% on SimpleQA with search, 74% on X Bench Deepsearch (Chinese), 66% on Reka Research Eval, 58% on X Browse, 51.2% on Chinese BrowseComp and 44.9% on BrowseComp. The scores are not directly comparable across benchmarks, but the consistent pattern is that retrieval and tool use are core to Grok’s proposition.
Average Vending-Bench net worth reported for Grok 4 across five runs.
Grok 3 Think score on AIME 2025 with xAI’s highest test-time compute setting.
The winning metric is not the biggest model, the biggest funding round or the biggest combined reach. It is whether distribution, compute and model capability translate into recurring usage at sustainable cost. Grok’s 2025 traffic dip and 2026 third-party compute deal show why utilization belongs beside capability on the scoreboard.
For buyers, this means testing real workflows rather than outsourcing procurement to leaderboard rank. Grok 4.6 is competitive on many 2026 agentic benchmarks, but not universally first. For builders, the important differentiators are its live X/web access, tool stack, context windows and cost structure. For investors, monetization and infrastructure utilization matter at least as much as valuation.
The same discipline applies to marketing comparisons. Traffic can indicate attention, app-store rank can indicate momentum, benchmark scores can indicate capability and financing can indicate strategic capacity—but none proves customer retention or unit economics on its own. Grok’s strongest 2026 case is the combination of those signals. Its weakest interpretation is to choose whichever metric is largest and use it as a proxy for everything else.
Three Charts That Explain Grok’s 2026 Position
These charts focus on the comparisons where visualization adds information: the shape of standalone traffic growth, the capital-raising cadence, and the newest agentic-model benchmark profile.
xAI disclosed financing events
Round/financing size in USD billions; debt + equity shown as one 2025 financing event.
Grok 4.6 agentic benchmark profile
Six percentage-based benchmarks from xAI’s August 2026 comparison.
The three charts tell one coherent story. Grok can generate enormous product attention, xAI can raise infrastructure-scale capital, and the newest models are genuinely competitive at frontier agentic work. The remaining question is whether those advantages compound into durable usage and revenue faster than the cost base grows.
10 Evidence-Based Decisions for Founders, Marketers and Operators
Each recommendation below points back to a specific statistic in this report. The goal is to turn Grok’s numbers into operating decisions rather than trivia.
xAI’s January 2026 disclosure is approximately 600M monthly active users across X and Grok; use Grok.com visits and app installs for product-specific adoption.
Grok.com grew 4,168% in February 2025 and another 293% in March, then declined through June before rebuilding to a December high.
Grok 4.5 averaged 15,954 output tokens per SWE Bench Pro task versus 67,020 for Opus 4.8 Max in xAI’s comparison—about 4.2× fewer.
Grok 4 Fast scored 95% on SimpleQA with search and 74% on X Bench Deepsearch (zh), making retrieval quality a workflow-level feature worth testing.
Colossus 1 exceeded 220K NVIDIA GPUs by May 2026, while SpaceXAI also opened capacity to Anthropic—evidence that utilization belongs beside raw GPU count.
Grok Business launched at $30 per seat per month, while Enterprise adds SSO, SCIM and advanced controls rather than forcing regulated teams into consumer tooling.
Voice Think Fast 2.0 reported 0.70s time-to-first-audio and costs $0.08 per audio minute, making latency and minute economics directly measurable.
Grok Imagine won 64.1% of xAI’s overall editing comparison against Runway Aleph, and the Imagine API supports up to 10 images per request.
Grok 4.5 launched July 16, 2026 and Grok 4.6 followed on August 12, so production evaluations need a refresh cadence measured in weeks, not years.
xAI completed a $20B Series E in January 2026 and was valued at a reported $250B in SpaceX’s February acquisition, alongside consumer, API, enterprise and compute businesses.
How This Report Was Built
This article is current as of August 26, 2026. We prioritized primary xAI/SpaceXAI disclosures for product benchmarks, API pricing, enterprise packaging, funding, Colossus infrastructure and government partnerships; Similarweb for the full 2025 grok.com traffic series; Sensor Tower for current Android app estimates; and Reuters for transaction terms and financing figures not fully disclosed by xAI. Statistics from 2023–2025 are retained only when they provide necessary trend context and are labeled by original year.
Sources used: SpaceXAI/xAI: Introducing Grok 4.6 (2026); SpaceXAI/xAI: Introducing Grok 4.5 (2026); xAI: Grok 4.1 Fast and Agent Tools API (2025); xAI: Grok 4 Fast (2025); xAI: Grok 4 (2025); xAI: Grok 3 Beta (2025); SpaceXAI: Grok Voice Think Fast 2.0 (2026); SpaceXAI: Voice Agent Builder (2026); xAI: Grok Imagine API (2026); SpaceXAI: Imagine API product page (2026); xAI: Grok Business and Grok Enterprise (2025); SpaceXAI: Colossus (2026); SpaceXAI: New Compute Partnership with Anthropic (2026); xAI: Series E funding announcement (2026); xAI: Series C funding announcement (2024); xAI: Series B funding announcement (2024); xAI: API Public Beta (2024); xAI: El Salvador nationwide AI education partnership (2025); xAI: U.S. DOW GenAI.mil partnership (2025); xAI: Saudi Arabia/HUMAIN partnership (2025); Similarweb: Winners and Losers in the Gen AI Market (2026); Sensor Tower: Grok Android App Overview (Aug. 2026); Reuters: SpaceX acquisition of xAI (Feb. 2026); Reuters/Morgan Stanley: xAI debt and equity financing (Jul. 2025); Reuters: xAI $10B raise / $200B valuation report (Sep. 2025); Reuters: xAI acquisition of X (Mar. 2025); Reuters: Telegram–xAI distribution agreement (May 2025); Reuters Breakingviews: Anthropic compute deal (May 2026); SpaceXAI: Company page (Aug. 2026). Where figures use different measurement bases, they are not combined. In particular, xAI’s ~600M MAU figure is preserved as combined X + Grok reach; Similarweb numbers are visits rather than unique users; Sensor Tower numbers are estimates; and benchmark scores are presented with their named evaluation rather than generalized into a universal ranking. When reporting xAI valuation, this article uses the value attached to the specific dated transaction/report instead of pretending private-company valuations are continuous audited market prices.