The AI Assistant Race, Late 2024 to October 2026: ChatGPT vs Claude vs Gemini vs Grok vs Meta AI vs DeepSeek, by the Numbers

Two years ago the AI assistant market looked like a one-horse race. In late 2024, ChatGPT took roughly six of every seven visits to generative-AI chatbot websites. Google's Gemini was a distant also-ran. Anthropic's Claude was a niche tool that developers liked, and DeepSeek was a name few people outside China had heard.

By October 2026 the picture has changed a lot, though not in the way most predictions said it would. ChatGPT is still by far the largest consumer AI product on Earth, and its user count has roughly quadrupled. But its share of the market has dropped steeply as Gemini rode Google's distribution and Claude grew from a rounding error into a serious consumer app. Meanwhile the money story ran almost the other way: on reported annualized revenue, Anthropic went from a fraction of OpenAI's size to matching or passing it, mostly by selling to businesses and software developers rather than to the general public.

This article pulls together the most credible public numbers we could find on users, web traffic, app downloads, revenue, enterprise market share, coding tools, funding and benchmark rankings. Every figure has a date and a source. Where sources disagree, we say so. Where a number is an estimate, a projection or comes from anonymous sources, we label it. At the end there is an honest look at who is ahead where, a release timeline, an FAQ, and a list of the sources we used.

A note on reading these numbers. AI companies report different metrics, often at times they choose. "Weekly active users", "monthly active users", "share of web visits", "share of downloads" and "annualized run-rate revenue" all measure different things. A run rate usually means the latest month's revenue multiplied by 12. It is not audited annual revenue, and it can swing a lot. Web-traffic shares from Similarweb cover websites only, not phone apps or features built into other products (such as Gemini inside Google Search, or Meta AI inside WhatsApp). Keep those limits in mind as you read the tables below.

The short version

• Consumer scale: ChatGPT leads by a wide margin. OpenAI reported 900 million weekly ChatGPT users in February 2026. At its September 29, 2026 DevDay it reportedly cited 1.2 billion weekly users. No rival has published a weekly number anywhere close.
• Share: ChatGPT's lead has shrunk. Similarweb's figures put ChatGPT at about 86.7% of generative-AI chatbot web traffic in late 2024. That fell to 52.7% in May 2026, then recovered a little to 56.4% in September 2026. Gemini went from roughly 5% to about 25%, and Claude from under 2% to about 9.6%.
• Revenue: Anthropic caught up fast. Anthropic's reported run rate went from about $1 billion in early 2025 to more than $65 billion by the end of July 2026 (the latter from anonymous sources, via Bloomberg). Reports on OpenAI's run rate in autumn 2026 range from about $50 billion to nearly $70 billion, depending on the outlet and the accounting.
• Enterprise and coding: Claude has been strong. In Menlo Ventures' mid-2025 survey, Anthropic held 32% of enterprise LLM API usage to OpenAI's 25%, and 42% of the coding segment. JetBrains' mid-2026 developer survey found Claude Code used at work by 39% of professional developers, ahead of GitHub Copilot (21%) and OpenAI Codex (16%).
• Model quality: there is no single leader. On the public Arena text leaderboard (October 27, 2026), Google's Gemini 4 models hold the top two spots, with Anthropic's Claude Opus 5.5 third. On the Arena WebDev board, Claude Opus 5.5 is first and OpenAI's GPT-6 Astra second.
• Money in: the scale is unprecedented. OpenAI closed a $122 billion round at an $852 billion valuation (March 2026). Anthropic raised $65 billion at $965 billion (May 2026). xAI was folded into SpaceX in February 2026 at an implied $250 billion.

How big is ChatGPT? Weekly users, 2024–2026

OpenAI is the only major lab that regularly publishes a weekly active user figure for its main consumer product. Here are the milestones it has announced, with who said them and when.

The last row needs care. We could not open an OpenAI primary page with the 1.2 billion figure. It comes from secondary reports of the DevDay keynote and OpenAI's recap, and FourWeekMBA notes the phrasing was "collective" weekly users, without a split between free and paid.

ChatGPT weekly active users (each █ ≈ 50 million)

Dec 2024  ██████                     300M
Feb 2025  ████████                   400M
Mar 2025  ██████████                 500M
Aug 2025  ██████████████             ~700M
Oct 2025  ████████████████           800M
Feb 2026  ██████████████████         900M
Sep 2026  ████████████████████████   1.2B (reported, DevDay)

Sources: Windows Central (Dec 2024); Engadget (Feb 2025); TechCrunch (Oct 2025, Feb 2026); FourWeekMBA and Fit to Prompt (Sep 2026).

For the mobile app alone, Sensor Tower's State of AI 2026 report says ChatGPT reached 1 billion monthly active users in May 2026, which it called the fastest mobile app to do so. Sensor Tower's MAU numbers are third-party estimates, not company disclosures.

What about everyone else's users?

Rivals report differently and less often, so a clean like-for-like table isn't possible. Here is what has been disclosed or credibly estimated.

Two warnings. First, Google's and Meta's numbers are company-reported, and neither company has published an exact definition of a "monthly active user". Meta AI's figure counts use across Facebook, Instagram, WhatsApp and Messenger, where the assistant is built in. That is a very different thing from someone choosing to open a standalone chatbot. Second, Sensor Tower's estimates (the 1.1 billion, 662 million and 245 million figures) are modeled, not reported, and other outlets citing Sensor Tower have given much lower Claude app-only figures. Treat them as rough sizes, not precise counts.

We could not find a credible, recent company-disclosed user number for Grok or DeepSeek, so we have left them out of this table rather than guess.

Web traffic share: the clearest picture of the shift

Similarweb tracks visits to the websites of the main chatbots and publishes each one's share of the total. This is the most consistent long-run series available, but it is websites only. It misses phone apps, and it misses AI built into other products.

All figures are Similarweb data as reported by the outlets named. Dates marked "~" are inferred from the reports' "X months ago" wording. A dash means the source gave no value. Meta AI is not in Similarweb's tracked set.

Similarweb share of gen-AI chatbot web visits, September 2026 (each █ ≈ 2 percentage points)

ChatGPT     ████████████████████████████  56.4%
Gemini      ████████████▌                 25.0%
Claude      ████▊                          9.6%
DeepSeek    █▋                             3.4%
Grok        █▏                             2.3%
Copilot     ▊                              1.5%
Perplexity  ▌                              0.9%

Source: Similarweb, via FourWeekMBA (October 2026).

Three things stand out.

1. Gemini's rise was the biggest single change. It went from about 5% to more than a quarter of web visits in roughly 18 months. Its fastest gains came between October 2025 (13.6%) and March 2026 (25.5%), a period that included the Gemini 3 launch in November 2025. Google can also push Gemini through Android, Chrome, Search and Workspace.
2. Claude's jump came mostly in 2026. Claude sat around 2% through January 2026, then roughly tripled to 6% by March and about 9–10% by the summer.
3. ChatGPT's slide has paused, at least for now. After bottoming at 52.7% in May 2026, ChatGPT's share edged back up to 56.4% by September. Similarweb's broader report also notes ChatGPT's absolute visit numbers stayed roughly flat while the overall market grew about 70% in a year. In other words, ChatGPT didn't shrink. The market grew around it.

DeepSeek's story is a spike and a fade. After its R1 model launched free on January 20, 2025, DeepSeek briefly became the top free app on the US iOS App Store (January 27, 2025, per Wikipedia), and its web share hit about 9% in early 2025. It has since settled around 3–4%.

App downloads and consumer spending

Phones are where much of the consumer race now happens. Two app-intelligence firms, Sensor Tower and Appfigures, publish estimates.

Share of downloads among six leading AI assistant apps (Sensor Tower, via Forbes, May 2026)

Download share, Q2 2026 to date (each █ ≈ 2 points)
ChatGPT  ███████████████████████▌  47%
Gemini   ███████████               22%
Claude   ███████                   14%

Appfigures adds that competitors combined have out-downloaded ChatGPT every month since September 2025, even though ChatGPT is still the single biggest app. ChatGPT alone had about 54.5 million downloads in March 2026.

Estimated consumer spending, April 2026 (Appfigures, App Store + Google Play, net)

ChatGPT still earns about twice as much from app stores as its main rivals combined. In January 2025 that ratio was about 18.5 to 1, so the gap has narrowed a lot. Gemini's low app-store figure may not reflect its true paid base. Google sells its AI plans as Google One subscriptions, and Appfigures does not say whether those are attributed to the Gemini app.

Sensor Tower also says Claude's average US mobile revenue per user rose from under $0.50 in September 2025 to $2.76 in May 2026. Claude's paying base grew quickly. Its app-store spend went from $11 million in January–April 2025 to $156 million in January–April 2026 (Appfigures). ChatGPT still earned more per download in April 2026 ($5.83 vs $4.06).

What caused Claude's 2026 consumer jump? At least one identifiable event played a part. In late February 2026, after Anthropic refused to drop its limits on mass domestic surveillance and fully autonomous weapons, Defense Secretary Pete Hegseth said the Pentagon would designate Anthropic a "supply chain risk". The designation was made formal on March 5, 2026 (CBS News). OpenAI announced its own deal with the Defense Department soon after. Forbes, citing Similarweb, reported that Claude app downloads rose 55% week-over-week in the week after the dispute broke, while ChatGPT downloads fell 5.6%. A burst of attention doesn't explain a sustained trend on its own, though. Claude's growth continued for months, which points to product factors too, including Claude Code and new model releases.

Revenue: OpenAI vs Anthropic

Neither company is public, so revenue figures come from company statements, investor updates and press reports, almost always as annualized run rates. Treat them as indicators, not audited results.

Anthropic reported run-rate revenue

OpenAI reported run-rate revenue, 2026

Anthropic run rate (each █ ≈ $2.5B)
Early 2025  ▍                            ~$1B
Aug 2025    ██                           >$5B
End 2025    ███▋                         ~$9B
Feb 2026    █████▋                       $14B
Apr 2026    ████████████                 >$30B
May 2026    ██████████████████▊          >$47B
Jul 2026    ██████████████████████████   >$65B (anonymous sources)

OpenAI run rate, 2026 reports (each █ ≈ $2.5B)
Aug 2026    ████████████████             >$40B
Oct 2026    ████████████████████         ~$50B (Bloomberg/FT)
Sep 2026    ████████████████████████████ ~$70B (Axios)

Why do the OpenAI figures disagree so much? Bloomberg Government's October 8 piece says the gap comes partly from investors trying to compare OpenAI's numbers directly with Anthropic's, including how sales made through cloud providers are counted. Anthropic sells a large share of its models through Amazon, Google Cloud and Microsoft Azure, and whether you book those sales at the gross or net amount makes a big difference. So when you see a headline saying "Anthropic passed OpenAI on revenue", read it as "on one way of counting, at one moment". Both companies are growing at a pace with no real precedent, and both burn very large sums on compute.

Coding is a big part of both companies' growth. Anthropic said in February 2026 that Claude Code alone had a run rate above $2.5 billion, more than double its level at the start of the year, and that enterprises made up over half of Claude Code revenue. In September 2025 it had put Claude Code at over $500 million. Bloomberg Government's summary of the Axios report says OpenAI's recent growth was also driven partly by its AI coding software, though it gives no figure.

Enterprise and API market share

The best-known public estimate of enterprise market share comes from the venture firm Menlo Ventures, which surveys enterprise technology buyers. Menlo has invested in Anthropic, which is worth knowing when you read its numbers.

Menlo Ventures, mid-2025 LLM market update (survey of June 30–July 10, 2025)

Enterprise LLM API usage share, mid-2025 (each █ ≈ 2 points)
Anthropic  ████████████████  32%
OpenAI     ████████████▌     25%
Google     ██████████        20%
Meta       ████▌              9%
DeepSeek   ▌                  1%

Menlo put total enterprise spending on model APIs at $8.4 billion by mid-2025, up from $3.5 billion in November 2024. It also found open-source models' share of enterprise workloads fell from 19% to 13% in six months.

Secondary coverage of a later, end-of-2025 Menlo report says Anthropic rose to about 40% of enterprise LLM API spend, with OpenAI at 27% and Google at 21%. We could not open that report directly, so treat those later figures as reported rather than verified.

Anthropic's own customer disclosures point the same way. In February 2026 it said more than 500 customers were spending over $1 million a year (up from about a dozen two years earlier), and that eight of the Fortune 10 were customers.

Coding tools: where Claude has its clearest lead

AI coding agents became a breakout category over these two years. Products like Anthropic's Claude Code, OpenAI's Codex, GitHub Copilot (Microsoft) and Cursor now write a large share of new code at many companies.

JetBrains Developer Ecosystem Survey 2026 (15,000+ professional developers, May–July 2026), via Gigazine

Coding tool used at work, mid-2026 (each █ ≈ 2 points)
Claude Code     ███████████████████▌  39%
GitHub Copilot  ██████████▌           21%
OpenAI Codex    ████████              16%

The same survey found 90% of professional developers use AI coding agents at least weekly and 68% use them daily. Awareness of Claude Code and Copilot was tied at 79%, and Codex awareness rose from 27% to 65% within 2026.

Anthropic also cited a SemiAnalysis estimate in February 2026 that Claude Code accounted for about 4% of all public GitHub commits worldwide.

Developers often use more than one tool, so these percentages don't add up to 100, and survey samples differ. The direction is consistent across sources, though: Claude Code went from almost nothing to the most-used coding agent in about 18 months, and Codex is the fastest-growing challenger.

How Anthropic tried to win, and the usage-limit backlash

The strategy: enterprise, developers and coding

Anthropic never tried to beat ChatGPT at being a free chatbot for everyone. Its public moves through 2025–2026 show a different bet:

• Sell to businesses and developers first. Claude is available on all three big clouds (AWS, Google Cloud and Azure). Anthropic has repeatedly highlighted large enterprise accounts in its funding announcements.
• Make coding the wedge. Claude Code fully launched in May 2025 and became the company's fastest-growing product. Anthropic leaned into long-running "agentic" work, meaning models that plan, use tools and fix their own errors over hours.
• Ship models fast. Between February and October 2026 Anthropic released Opus 4.6, Sonnet 4.6, Opus 4.7, Opus 4.8, Fable 5, Sonnet 5, Opus 5, Fable 5.1, Opus 5.5, Sonnet 5.5 and Haiku 5.5 (see the timeline below).
• Buy a lot of compute. Anthropic's May 6, 2026 post lists deals with Amazon (up to 5 GW), Google and Broadcom (5 GW from 2027), Microsoft and Nvidia ($30 billion of Azure capacity), Fluidstack, and the entire Colossus 1 data center from SpaceX (300+ MW, 220,000+ Nvidia GPUs).
• Make safety part of the brand. That includes the public stand against the Pentagon over surveillance and autonomous weapons, which cost it government business but appears to have won consumer goodwill.

The backlash: usage limits on Pro and Max

The same demand that drove Anthropic's revenue also strained its capacity, and paying subscribers felt it. Here is a timeline of the changes we could verify.

Plenty of further claims circulated, such as a new tokenizer quietly shrinking allowances or exact percentages of "silent" limit cuts. Some come from vendors selling alternatives to Claude, and we couldn't verify them against primary sources, so we've left them out. As of July 2026, a third-party plan tracker reconfirmed Claude Pro at roughly 45 messages per 5 hours plus a weekly cap. Anthropic's limits are usage-based and depend on model, context length and time of day, so exact numbers vary.

To be fair to all sides: the complaints were real and widespread. Anthropic's communication in March 2026 was mostly informal (posts on X by staff, not an official notice), and $200-a-month subscribers had a reasonable grievance about that. On the other hand, Anthropic was serving demand that reportedly grew several times over in a few months, and it did reverse the most painful change within about six weeks once new compute arrived.

How OpenAI and Google handled limits

Anthropic wasn't alone. Every lab selling flat-rate plans for agentic coding has wrestled with the same economics.

OpenAI

Google. On May 15, 2026, according to a third-party plan tracker citing Google's support pages, Google replaced fixed daily prompt counts for Gemini (5 a day free, 100 a day on AI Pro, 500 a day on Ultra) with compute-based allowances that refresh every 5 hours within a weekly cap. This changed how limits are measured. We could not confirm that it reduced them.

The pattern is the same at all three: flat-price plans built for chat struggle when users run agents for hours, so companies move to rolling windows, weekly caps and pricier tiers. Anthropic's version drew the loudest backlash partly because its users are heavy coding users, and partly because of how it communicated the changes.

Did the backlash hurt Claude, or help ChatGPT?

Short answer: there is no clear evidence in the public data that it did, at least not yet. Here's what the numbers show:

• Claude's web share kept rising through the worst of it. Similarweb had Claude at about 6.0% around March 2026, 8.0% in April and 8.9% in May, all during or right after the limit complaints. It reached 9.6% in July.
• Claude's app download share rose too. Sensor Tower's Q2 2026 quarter-to-date figure (14%) covers the months after the March–April backlash.
• Claude's growth has since flattened. Its web share sat at 9.6%, 9.3% and 9.6% in July, August and September 2026.
• ChatGPT did regain some share. It went from 52.7% in May to 56.4% in September 2026.

Could the flattening and ChatGPT's partial rebound be linked to the limits? Possibly, but other things happened over the same months. OpenAI shipped GPT-5.6 and GPT-6, and Google's Gemini share also slipped (27.3% to 25.0%). Nobody has published subscriber churn data for Claude. Reddit threads and canceled-subscription posts show real anger among power users, but they don't measure overall behavior. Based on what is public, we can't say the backlash measurably hurt Claude or helped ChatGPT.

Funding and valuations, 2024–2026

The amounts of money involved grew roughly tenfold in two years.

Google DeepMind doesn't raise outside money. It is funded by Alphabet. The closest comparable figure is Alphabet's capital spending: in late July 2026 it raised its 2026 capex guidance to $195–205 billion, up from $180–190 billion, driven by AI infrastructure demand (W.Media, July 23, 2026). That is spending on data centers and chips, not a valuation, but it shows how much Google can put behind Gemini.

Latest private valuation (each █ ≈ $40B)
Anthropic (May 2026)  ████████████████████████   $965B
OpenAI    (Mar 2026)  █████████████████████▎     $852B
xAI       (Feb 2026)  ██████▎                    ~$250B (implied, in SpaceX merger)

Valuation path (each █ ≈ $40B)
OpenAI     Oct 2024 ████ $157B → Mar 2025 ███████▌ $300B → Oct 2025 ████████████▌ $500B → Mar 2026 █████████████████████▎ $852B
Anthropic  Mar 2025 █▌ $61.5B → Sep 2025 ████▌ $183B → Feb 2026 █████████▌ $380B → May 2026 ████████████████████████ $965B
xAI        Jan 2026 █████▊ $230B → Feb 2026 ██████▎ $250B

Alphabet 2026 capex guidance (not a valuation): █████ ~$200B

Two cautions. Headline round sizes are often paid in tranches tied to conditions (SoftBank's $40 billion in 2025 depended on OpenAI's restructuring). And private valuations are set by a small number of investors, so they can move fast in either direction. Secondary-market prices quoted on aggregator sites in October 2026 put OpenAI above $1 trillion, but those are not formal funding rounds.

The models: how the frontier compares in late 2026

Benchmark leaderboards

The most-watched public ranking is Arena (formerly LMArena, at arena.ai). Real users compare two anonymous models' answers side by side and vote, and the results are turned into Elo-style scores. It measures what people prefer, which overlaps with capability but isn't the same thing.

Arena Text leaderboard, overall (snapshot dated October 27, 2026)

Arena WebDev leaderboard (snapshot dated October 27, 2026)

Arena Text, Oct 27 2026 (bars show points above 1480; each █ ≈ 2 points)
Gemini 4 Carbon   ████████████████████████  1528
Gemini 4 Argon    ████████████████████      1520
Claude Opus 5.5   ██████████████▌           1509
Muse Wonder       █████████████▌            1507
Claude Fable 5.1  ████████████              1504

The small gaps matter. Scores within about 10 points of each other often fall inside the confidence intervals, so the top five text models are close to a statistical tie. One notable detail: no OpenAI model appears in the top 15 of the text board in this snapshot, but OpenAI holds two of the top five WebDev spots.

Other independent measures

The two Vals figures for Opus 5.5 differ (70.18% vs 66.97%), probably because the index's test set or run settings changed between snapshots. That's a useful reminder that benchmark numbers are a moving target.

Vendor-reported coding scores (treat with caution)

Each company picks the benchmarks and settings it publishes, so vendor numbers are not a neutral comparison.

Reception of the newest models

• OpenAI's GPT-6 family. GPT-6 Astra (flagship) arrived September 3, 2026, followed by GPT-6 Sol and GPT-6 Luna on September 22 and GPT-6.1 Sol at DevDay on September 29. Astra ranks second on Arena WebDev and does well on business-automation and computer-use tasks. Independent indexes place it behind Anthropic's latest on many developer benchmarks, and Artificial Analysis measured a high hallucination rate for it. Its API list price ($10 input / $50 output per million tokens, rising for very long prompts, per Gradually.ai) is higher than Opus 5.5's.
• Anthropic's Claude 5 family. Fable 5.1 (released September 1–2, 2026; coverage differs by a day) is built on the same base as the restricted Mythos 5.1, with tighter safety settings. It is aimed at multi-hour agentic and scientific work. Anthropic claims big gains on its own Terminal-Bench tests (52.6% on Terminal-Bench-Science 0.1, versus 22.4% it reported for GPT-5.6 Sol) and estimates about 25% lower costs for typical work, up to 45% for agent-heavy work. Early independent tests were mixed: in some, Fable 5.1 cost about 20% more than Fable 5 because it produced more output tokens, while one agent-heavy test cut cost nearly in half (Tecnoblog). It is priced at $10 / $50 per million tokens. Opus 5.5 (September 22, 2026) costs $4 / $20 per million tokens, 20% below Opus 5's list price. Anthropic says typical workloads cost up to 40% less than on Opus 5 and that it reaches Fable 5.1's level on most tasks. It tops Arena WebDev and the Artificial Analysis index. Sonnet 5.5 (September 28) costs $2 / $10 and scores close to Opus 5.5 on many tests. Haiku 5.5 followed on October 7. Some cybersecurity requests are still routed to older Opus models for safety reasons, which some users find limiting.
• Google's Gemini 4. Gemini 4 Argon was announced September 30, 2026 for approved cyber-defense partners, then opened to developers, enterprises and paid subscribers on October 13. It ranks first or second on several independent boards and has a notably low measured hallucination rate. Gemini 4 Carbon appears at the top of Arena's text board, but at the time of writing it is listed as a limited preview.
• Meta's Muse. Meta Superintelligence Labs released Muse Spark on April 8, 2026 (free inside Meta's apps), then the Muse Wonder and Wonder Full frontier models on October 12, 2026. Wonder Full scores 56 on the Artificial Analysis index, which puts Meta back among the frontier labs on that measure. Meta has published few details about how the models were built.
• xAI's Grok. xAI kept shipping point releases, Grok 4.5 through 4.7, between July and September 2026. Grok 5 had not been released as of the latest sources we found. Grok has also faced repeated controversies over content and safety, documented on Wikipedia, which matters for enterprise buyers.
• DeepSeek. DeepSeek released V4 previews in April 2026, then V4-Flash (July 31), V4-Pro (August 13) and V4.1-Flash (September 10). Its open-weight models stay popular with developers on a budget, but its web share hasn't returned to its early-2025 peak.

Can ChatGPT beat Claude, and where is each one ahead?

The question assumes a single race. The data shows at least three, and different companies are winning each one.

Where ChatGPT is clearly ahead

Consumer scale. ChatGPT has more weekly users (900 million confirmed in February 2026, a reported 1.2 billion by late September) than Claude is estimated to have monthly users, by a factor of several. Even on Sensor Tower's generous estimate of 245 million Claude MAU, ChatGPT is roughly four to five times bigger. ChatGPT still has more than half of chatbot web traffic, earns about twice as much from app stores as its main rivals combined, and has the strongest consumer brand. "ChatGPT" has become a generic word for AI chat in much of the world.

Breadth of product. ChatGPT bundles chat, image generation, voice, agents ("Dots", announced at DevDay 2026), a developer platform, and a growing set of apps inside it. OpenAI is pitching ChatGPT itself as a platform where other companies build. That's a distribution advantage no pure model company can easily copy.

So: can Claude beat ChatGPT for consumers? On current trends, not soon. Claude's web share has stalled around 9–10% since July 2026, while ChatGPT's has edged up.

Where Claude has been strong

Coding and agents. This is Claude's clearest lead. It is the most-used coding agent among professional developers in JetBrains' mid-2026 survey, by a wide margin. It tops the Arena WebDev board, and Opus 5.5 leads the Artificial Analysis index and most of Vals AI's head-to-head benchmarks against GPT-6 Astra.

Enterprise API. Menlo's surveys put Anthropic ahead of OpenAI in enterprise API usage since mid-2025. Anthropic's revenue, much of it from business customers, has grown faster than OpenAI's on most reported measures.

Fast-growing paid base. Claude's app-store spending grew roughly fourteenfold year over year in early 2026 (Appfigures), and Sensor Tower says its US mobile revenue per user rose from under $0.50 to $2.76 between September 2025 and May 2026.

So: can ChatGPT beat Claude in enterprise and coding? It's plausible. Codex grew from zero to 16% workplace usage in about 18 months, GPT-6 Astra is a strong second on WebDev, and OpenAI has the bigger consumer funnel feeding business adoption. Enterprise buyers also switch models more readily than consumers do, since many use several providers through the same cloud platforms. Claude's lead in coding is real, but it isn't locked in.

And Google, the quiet third front-runner

Any honest answer has to include Gemini. Google tops the Arena text board as of late October 2026, has about a quarter of chatbot web traffic, reports more than a billion monthly Gemini app users, and can spend roughly $200 billion a year on capex. Gemini also reaches people through Android, Search, Chrome and Workspace without them ever visiting a chatbot website. If the question is "who is most likely to take consumer share from ChatGPT?", the data points to Google more than to Anthropic.

What actually decides the race

1. Compute. Every lab is limited by chips and power. Anthropic's 2026 usage-limit crisis was a compute shortage made visible, and it eased when the SpaceX Colossus 1 deal came online. OpenAI's Stargate project, Google's in-house TPUs and Alphabet's capex, and xAI's position inside SpaceX are all ways of solving the same problem. Whoever has spare capacity can offer generous plans. Whoever runs short has to ration.

2. Distribution. Google and Meta can put an assistant in front of billions of people at almost no cost to acquire them. OpenAI built distribution by being first and becoming the default name. Anthropic relies on developers, enterprises and cloud partners. Distribution explains why Gemini and Meta AI can report huge user counts while Claude's consumer numbers stay smaller.

3. Price. Frontier API prices keep falling. Opus 5.5 came in cheaper than Opus 5, Gemini 4 Argon launched at an introductory $2/$10, and DeepSeek's open models undercut everyone. At the same time, consumer plans are getting more expensive at the top ($200–$500 a month tiers). The labs are moving toward charging for actual compute used rather than flat fees.

4. Model quality. The frontier is crowded. Google, Anthropic, Meta and OpenAI all have models within a few Arena points of each other, and leads change within weeks. Model quality is necessary to compete but doesn't settle things, since a lead rarely lasts a full quarter.

5. Agents. The next battleground is models that do multi-hour work: writing and testing code, running business workflows, operating computers. That favors labs with strong coding models (Anthropic), strong product platforms (OpenAI), and integrated ecosystems (Google). Agent workloads also use far more compute per user, which links back to point 1.

6. Trust and safety. This one cuts several ways. Anthropic's stand against the Pentagon cost it government business but seems to have won consumer goodwill. Its usage-limit communication in March 2026 cost it goodwill among its most loyal users. Grok's repeated content controversies are a known obstacle for enterprise sales. Measured hallucination rates differ widely between top models. Enterprises in regulated industries increasingly weigh reliability and data handling as heavily as raw benchmark scores.

Our bottom line

• ChatGPT is the consumer leader and will likely stay that way for the foreseeable future, though with a smaller share of a much bigger market.
• Claude has led coding and much of enterprise API usage, and Anthropic has become a revenue peer of OpenAI. But its consumer growth has flattened, and it has to manage capacity better than it did in early 2026.
• Gemini is the most serious threat to ChatGPT's consumer lead, and Google currently holds the top text-leaderboard spot.
• Meta, xAI and DeepSeek are each competitive in parts of the market (distribution inside social apps, X integration, and low-cost open weights respectively), but none leads a major segment on the data available.

Nobody has "won". The market is splitting by use case, and many buyers now use several models at once.

How small builders play the field: Artificial Mind Hive

Disclosure: Apiary and Artificial Mind Hive share an owner.

The data above describes the giants. One of the clearest trends in 2026, though, is happening at the small end. Many independent builders no longer pick one AI company. They mix models by task and by cost.

Artificial Mind Hive is one example. It is a small independent studio that runs a multi-agent setup instead of betting on a single lab. Founder Austin Little runs Claude, Codex, Grok and local models side by side, each in its own lane, and sends work to whichever does the job best for the cost:

• Claude (Anthropic) handles user-interface and website work.
• OpenAI's Codex handles video and data jobs.
• xAI's Grok handles hub and operations tasks.
• Local open models run through Ollama handle cheap, private work that doesn't need to leave the studio's own machines.

The studio also publishes an "Agent Book" at amhagentbook.com, where outside AI agents can join over MCP (the Model Context Protocol, an open standard for connecting AI agents to tools and services).

This setup mirrors the pattern in the market data. Developers report using more than one coding tool, enterprises spread API spending across several providers, and leaderboard leads change monthly. For a small team, routing each job to the model that does it best for the cost hedges against outages, usage limits and price changes, all of which hit users of every major lab in 2025–2026. The usage-limit episodes covered above are exactly the kind of disruption a multi-model setup is meant to absorb.

Timeline of major model releases, late 2024 to October 2026

Dates come from Presenc.ai (OpenAI), a release timeline by Hidekazu Konishi (Anthropic), Wikipedia (Gemini, Grok, DeepSeek), and launch coverage cited in the sources. Where sources differ by a day, we show a range.

FAQ

Which AI chatbot has the most users in 2026? ChatGPT. OpenAI reported 900 million weekly active users in February 2026 and reportedly cited 1.2 billion weekly users at its September 29, 2026 DevDay. Google reported more than 1 billion monthly Gemini app users in August 2026, and Meta reported about 1 billion monthly Meta AI users back in May 2025. Weekly and monthly figures aren't directly comparable, and Meta's figure counts an assistant built into apps people already use.

Is ChatGPT losing market share? Yes in share, no in absolute terms. Similarweb's data shows ChatGPT's share of chatbot web visits falling from about 86.7% in late 2024 to 52.7% in May 2026, then recovering to 56.4% by September 2026. Its user numbers kept growing over the same period. The market grew faster than ChatGPT did.

Has Anthropic really passed OpenAI on revenue? On some reported measures, at some points in 2026, yes. Anthropic said its run rate passed $30 billion in April 2026, and unnamed sources told Bloomberg it was above $65 billion by the end of July. Reports on OpenAI in autumn 2026 range from about $50 billion to nearly $70 billion. The companies count revenue sold through cloud partners differently, and run rates are not audited. A fair summary is that the two are now in the same range.

Why is Claude so popular with programmers? Anthropic focused on coding and long-running agent tasks early. Claude Code launched fully in May 2025 and by mid-2026 was used at work by 39% of professional developers in JetBrains' survey. Claude models have also ranked at or near the top of coding-focused leaderboards such as Arena WebDev.

What were the Claude usage-limit complaints about? Anthropic added weekly caps to Pro and Max in August 2025. In March 2026 it made 5-hour session limits drain faster during weekday peak hours, and users also reported possible caching bugs. Many paying subscribers complained, especially about the lack of clear official communication. On May 6, 2026, after adding SpaceX compute, Anthropic doubled Claude Code's 5-hour limits and removed the peak-hour reduction. Weekly caps stayed in place.

Do OpenAI and Google limit usage too? Yes. OpenAI restored a 5-hour cap on Plus for Codex and Work on August 25, 2026, added $100 and $500 Pro tiers in 2026, and reduced the allowance on new $200 Pro subscriptions after DevDay 2026. Google switched Gemini from daily prompt counts to compute-based limits on 5-hour and weekly windows in May 2026.

Which AI model is the "best" right now? It depends on the task. On Arena's text leaderboard (October 27, 2026), Google's Gemini 4 Carbon and Argon rank first and second, with Claude Opus 5.5 third. On Arena WebDev, Claude Opus 5.5 is first and GPT-6 Astra second. On the Artificial Analysis index, Claude Opus 5.5 leads. Leads at the top often change within weeks.

What happened to DeepSeek? DeepSeek's free R1 model caused a sensation in January 2025, briefly topping the US App Store and coinciding with an 18% one-day drop in Nvidia's share price. Its web share peaked at about 9% in early 2025 and has since settled around 3–4%. It keeps releasing open-weight models (V4 in 2026) that are popular with cost-conscious developers.

Where does Grok fit in? Grok holds about 2–4% of chatbot web traffic per Similarweb, plus distribution inside X. Its app earned an estimated $39 million in consumer spending in April 2026 (Appfigures). xAI was merged into SpaceX in February 2026 at an implied $250 billion valuation. Grok 5 had not shipped as of the latest sources we found.

How much money has been invested in these companies? Tens of billions per round. OpenAI closed a $122 billion round at an $852 billion valuation in March 2026. Anthropic raised $65 billion at a $965 billion valuation in May 2026. xAI raised $20 billion in January 2026 before its merger with SpaceX. Google funds Gemini internally, and Alphabet guided to $195–205 billion of capital spending in 2026.

Is it smart to rely on just one AI company? More and more builders don't. Usage limits, price changes, outages and fast-changing leaderboards have pushed many developers and businesses to use several models and send each task to the one that fits best on quality and cost.

Sources and further reading

Users and traffic

• TechCrunch, "Sam Altman says ChatGPT has hit 800M weekly active users" (Oct 6, 2025): https://techcrunch.com/2025/10/06/sam-altman-says-chatgpt-has-hit-800m-weekly-active-users/
• TechCrunch, "ChatGPT reaches 900M weekly active users" (Feb 27, 2026): https://techcrunch.com/2026/02/27/chatgpt-reaches-900m-weekly-active-users
• Engadget, "ChatGPT reaches 400M weekly active users" (Feb 20, 2025): https://engadget.com/ai/chatgpt-reaches-400m-weekly-active-users-203635884.html
• Windows Central, "ChatGPT surpasses 300 million weekly users" (Dec 2024): https://www.windowscentral.com/software-apps/chatgpt-surpasses-300-million-weekly-users
• FourWeekMBA, "ChatGPT's 1.2B Weekly Users, and Where DevDay Landed" (Sep–Oct 2026): https://fourweekmba.com/ai-chatgpt-12b-weekly-users-devday-2026-distribution/
• Fit to Prompt newsletter, DevDay 2026 issue (Sep 30, 2026): https://buttondown.com/fittoprompt/archive/openai-devday-brings-always-on-dots-agents/
• Implicator, "Google puts Gemini at 1 billion monthly users…" (Aug 2026): https://www.implicator.ai/google-puts-gemini-at-1-billion-monthly-users-without-naming-paying-subscribers/
• NBC San Diego / CNBC, "Mark Zuckerberg says Meta AI has 1 billion monthly active users" (May 28, 2025): https://www.nbcsandiego.com/news/business/money-report/mark-zuckerberg-says-meta-ai-has-1-billion-monthly-active-users/3835514/?amp=1
• FourWeekMBA, "Claude's AI Web Traffic Share Quadruples in a Year to 9.6%" (Oct 2026, Similarweb data): https://fourweekmba.com/ai-claudes-ai-web-traffic-share-quadruples-in-a-year-to-9-6/
• PPC Land, "ChatGPT drops to 52.7% as Claude triples its AI traffic share" (Jun 2026): https://ppc.land/chatgpt-drops-to-52-7-as-claude-triples-its-ai-traffic-share
• Similarweb, "AI Search Stats 2026": https://aisearch.similarweb.com/blog/gen-ai-stats/
• The Decoder, "ChatGPT's web traffic share dropped from 78% to 54% in one year…" (May 14, 2026): https://the-decoder.com/chatgpts-web-traffic-share-dropped-from-78-to-54-in-one-year-as-gemini-quietly-tripled-its-reach/
• Android Headlines, AI traffic share report update (May 8, 2025): https://www.androidheadlines.com/2025/05/ai-traffic-share-report-update-chatgpt.html

Apps and consumer spending

• Sensor Tower, State of AI 2026 press release: https://sensortower.com/press/sensor-tower-state-of-ai-2026-report-global-time-spent-on-generative-ai-apps-projected-to-more-than-double-year-over-year
• MediaNama, "ChatGPT's AI market share drops below 50%…" (Jun 17, 2026): https://www.medianama.com/2026/06/223-chatgpts-ai-market-share-drops-below-50-gemini-claude-gain-users-report/
• Forbes Australia, "ChatGPT… is losing downloads as Claude surges" (May 2026): https://www.forbes.com.au/news/innovation/chatgpt-the-fastest-growing-app-ever-is-losing-downloads-as-claude-surges/
• Appfigures, "ChatGPT at 3: Still Earning 2x Its Rivals Combined": https://appfigures.com/resources/insights/chatgpt-ai-chatbot-revenue-rivals

Revenue, funding and valuations

• The Decoder, "Anthropic increases revenue sevenfold, hits annualized rate above $65 billion" (Aug 18, 2026): https://the-decoder.com/anthropic-increases-revenue-sevenfold-hits-annualized-rate-above-65-billion/
• Simon Willison, notes on Anthropic run-rate disclosures (May 29, 2026): https://simonwillison.net/2026/May/29/anthropic/
• Anthropic, Series E announcement (Mar 3, 2025): https://www.anthropic.com/news/anthropic-raises-series-e-at-usd61-5b-post-money-valuation
• Anthropic, Series F announcement (Sep 2, 2025): https://www.anthropic.com/news/anthropic-raises-series-f-at-usd183b-post-money-valuation
• Anthropic, Series G announcement (Feb 12, 2026): https://www.anthropic.com/news/anthropic-raises-30-billion-series-g-funding-380-billion-post-money-valuation
• Built In, "Anthropic Raises $65B Series H, Achieves $965B Valuation" (Jun 1, 2026): https://builtin.com/articles/anthropic-raises-65b-series-h-965b-valuation-20260601
• Built In, "OpenAI Raises $122B in Funding at $852B Valuation" (Apr 1, 2026): https://builtin.com/articles/openai-raises-122b-852b-valuation-20260401
• SiliconANGLE, "OpenAI bags $40B funding…" (Mar 31, 2025): https://siliconangle.com/2025/03/31/openai-bags-40b-funding-increasing-post-money-valuation-300b
• Wikipedia, "OpenAI": https://en.wikipedia.org/wiki/OpenAI
• Bloomberg Government, "OpenAI's Revenue Run Rate Nears $50 Billion, Less Than Reported" (Oct 8, 2026): https://news.bgov.com/artificial-intelligence/openais-annualized-revenue-nears-50-billion-ft-says
• Bloomberg Government, "OpenAI's Annualized Revenue Nears $70 Billion, Axios Says" (Sep 2026): https://news.bgov.com/artificial-intelligence/openais-annualized-revenue-nears-70-billion-axios-says
• Implicator, "SpaceX acquires xAI in $1.25 trillion deal…" (Feb 2, 2026): https://www.implicator.ai/spacex-acquires-xai-in-1-25-trillion-deal-creating-worlds-most-valuable-private-company/
• Longterm Wiki, xAI funding history: https://www.longtermwiki.com/organizations/xai/funding
• W.Media, "AI infrastructure demand pushes Alphabet's 2026 capex guidance to US$205 billion" (Jul 23, 2026): https://w.media/ai-infrastructure-demand-pushes-alphabets-2026-capex-guidance-to-us-205-billion/

Enterprise and coding

• Menlo Ventures, "2025 Mid-Year LLM Market Update": https://menlovc.com/insights/2025-mid-year-llm-market-update
• Gigazine, AI coding agent adoption trends / JetBrains survey (Aug 24, 2026): https://gigazine.net/gsc_news/en/20260824-ai-coding-agent-adoption-trends/

Usage limits and policy

• OfficeChai, "Anthropic Introduces Weekly Rate Limits For Claude Pro and Max Subscribers" (Jul 29, 2025): https://officechai.com/ai/anthropic-introduces-weekly-rate-limits-for-claude-pro-and-max-subscribers
• PiunikaWeb, "Anthropic explains Claude usage limits, peak hours" (Mar 27, 2026): https://piunikaweb.com/2026/03/27/anthropic-explains-claude-usage-limits-peak-hours/
• PiunikaWeb, "Anthropic Claude usage limits apology backlash" (Apr 3, 2026): https://piunikaweb.com/2026/04/03/anthropic-claude-usage-limits-apology-backlash/
• Anthropic, "Higher limits and a compute deal with SpaceX" (May 6, 2026): https://anthropic.com/news/higher-limits-spacex
• DEV Community, "Why Claude Max Users Are Leaving in May 2026" (opinion piece from a vendor of alternatives; read for context, not data): https://dev.to/owen_fox/why-claude-max-users-are-leaving-in-may-2026-a-data-driven-look-at-the-throttling-backlash-159p
• Business Today, OpenAI's new $100 ChatGPT Pro plan (Apr 10, 2026): https://www.businesstoday.in/amp/technology/story/openai-introduces-new-chatgpt-pro-plan-check-price-features-and-upgrades-525099-2026-04-10
• Notebookcheck, "OpenAI abruptly restores harsh 5-hour Codex and Work limits for ChatGPT Plus" (Aug 2026): https://www.notebookcheck.net/OpenAI-abruptly-restores-harsh-5-hour-Codex-and-Work-limits-for-ChatGPT-Plus.1378377.0.html
• ProPakistani, "ChatGPT Gets $500 Pro Plan as OpenAI Halves Pro 200 Limits" (Sep 30, 2026): https://propakistani.pk/2026/09/30/chatgpt-gets-500-pro-plan-as-openai-halves-pro-200-limits/
• AI Plan Finder, AI limits tracker: https://aiplanfinder.com/
• CBS News, Pentagon–Anthropic supply chain risk dispute (Mar 2026): https://www.cbsnews.com/news/pentagon-anthropic-supply-chain-risk-feud-ai-guardrails

Models and benchmarks

• Arena text leaderboard (snapshot Oct 27, 2026): https://arena.ai/leaderboard/text
• Arena WebDev leaderboard (snapshot Oct 27, 2026): https://arena.ai/leaderboard/webdev
• Vals AI, Claude Opus 5.5 vs GPT-6 Astra comparison: https://www.vals.ai/comparisons/anthropic_claude-opus-5-5-vs-openai_gpt-6-astra
• Let's Data Science, "Gemini 4 Argon: public independent tests rank it first and eighth" (Oct 2026): https://letsdatascience.com/blog/gemini-4-argon-public-independent-tests-rank-it-first-and-eighth
• Mezha, Anthropic Claude Fable 5.1 launch coverage (Sep 2026): https://mezha.ua/en/news/anthropic-claude-fable-5-1-314761/amp/
• Tecnoblog, "Anthropic lança Fable 5.1…" (Sep 2026): https://tecnoblog.net/noticias/anthropic-lanca-fable-5-1-com-menos-restricoes-e-promessa-de-economia/
• TechNews (Taiwan), Claude Opus 5.5 launch (Sep 23, 2026): https://technews.tw/2026/09/23/anthropic-upgrades-claude-with-new-opus-5-5-model/
• Gradually.ai, Claude Opus 5.5 vs GPT-6 Astra pricing comparison: https://www.gradually.ai/en/llm-comparison/claude-opus-5.5-vs-gpt-6-astra/
• Let's Data Science, "Meta launches Muse Wonder and Muse Wonder Full" (Oct 2026): https://letsdatascience.com/news/meta-launches-muse-wonder-and-muse-wonder-full-for-agentic-t-41113cca
• Presenc.ai, OpenAI GPT model lineage 2026: https://presenc.ai/research/openai-gpt-model-lineage-2026
• Hidekazu Konishi, Anthropic Claude model release timeline: https://hidekazu-konishi.com/entry/anthropic_claude_model_release_timeline.html
• Wikipedia, "Gemini (language model)": https://en.wikipedia.org/wiki/Gemini_(language_model)
• Wikipedia, "Grok (chatbot)": https://en.wikipedia.org/wiki/Grok_(chatbot)
• Wikipedia, "DeepSeek": https://en.wikipedia.org/wiki/DeepSeek