Codex Hits 4M Weekly Developers: The Growth Data

OpenAI Codex now reaches more than 4 million developers every week — a figure first disclosed on April 21, 2026 and re-confirmed today in the Dell Technologies on-prem partnership announcement. That trajectory, from roughly 600K weekly active users at the start of 2026 to 4M+ in under five months, represents a 6.7x growth multiple that no other coding-agent vendor has matched in publicly disclosed metrics.

The headline number matters less than the disclosure pattern behind it. OpenAI has published six distinct WAU milestones in five months — a cadence no other major coding-agent vendor maintains. Anthropic discloses growth rates without absolute counts, Cursor reports revenue rather than users, and GitHub Copilot surfaces subscriber numbers through Microsoft earnings calls. Codex is alone in making week-by-week user counts a part of its public communications strategy.

This post traces the full WAU trajectory, interrogates the denominator problem that makes cross-vendor comparisons unreliable, examines the enterprise engine behind Q1 2026's $5.7B OpenAI revenue quarter, and frames what today's Dell partnership signals about the next phase of growth. For context on what 4M developers are actually building, see our Codex "almost everything" release guide.

The 4M weekly active developer count was first disclosed in OpenAI's April 21 enterprise scaling announcement: “In early April, we shared that more than 3 million developers were using Codex every week. Just two weeks later, that number has grown to more than 4 million.” That same figure appeared again in the May 14 mobile launch post and is quoted directly in today's Dell partnership announcement: “More than 4 million developers now use Codex every week, and companies are already using it across the software development lifecycle, from code review and test coverage to incident response and reasoning across large repositories.”

The three-peat of the same number across a 27-day window is itself a data point. Either growth has plateaued near the 4M mark, or OpenAI's disclosure cadence trails the actual trajectory by several weeks. Both readings are plausible. What is clear is that 4M is the current public anchor — every comparison and competitive analysis should use this figure rather than round-number extrapolations.

The April 21 announcement also introduced Codex Labs — a partnership program with Accenture, Capgemini, CGI, Cognizant, Infosys, PwC, and Tata Consultancy Services — and named specific enterprise customers including Virgin Atlantic, Ramp, Notion, Cisco, and Rakuten. Those names are vendor-curated, but they establish the breadth of industry coverage behind the WAU number.

The WAU growth curve is unusually well-documented for a product in hypergrowth. OpenAI published or confirmed six milestones between January and April 2026, each tied to a product launch or announcement that gives the numbers additional credibility beyond press-release padding.

The fastest segment was the final two weeks before April 21: Codex added roughly 1 million WAU between the 3M disclosure on April 8 and the 4M disclosure on April 21 — an implied run-rate of approximately 70,000 new weekly active developers per day. The slowest growth appeared in the mid-March to early April window (2M to 3M took roughly three weeks), though that comparison is compressed by the uneven reporting cadence.

One important qualifier: the January 2026 baseline of ~600K comes from aggregator site Getpanto citing OpenAI public disclosures. OpenAI has not, as of publication, linked a primary announcement to the 600K figure directly. If that baseline proves higher, the growth multiple shrinks proportionally. We treat 6.7x as the defensible floor estimate, with the caveat noted.

The growth acceleration between February and April suggests that enterprise adoption — not consumer expansion — was the primary driver. The 1.6M Fortune figure on March 4 specifically called out Cisco, Nvidia, Ramp, Rakuten, and Harvey as early deployers. Enterprise onboarding tends to be lumpy: large deployment agreements add significant WAU in single steps, rather than the smoother ramp typical of consumer products.

For the broader competitive landscape including Cursor and Claude Code data, see our April 2026 Cursor / Copilot / Claude competitive snapshot.

The most analytically significant aspect of the 4M figure is not the number itself — it is the fact that OpenAI publishes it at all, repeatedly, with dates and product context attached. Across the major coding-agent vendors, Codex is the only product maintaining what amounts to a quarterly WAU disclosure cadence.

This is a deliberate competitive-signaling choice. Publishing absolute user counts on a recurring basis creates a trackable growth narrative that compounds over time. Each milestone announcement generates media coverage, and each subsequent milestone is framed against the previous one — turning a single data point into a growth story. The cadence itself is the asset.

Anthropic's approach is instructively different. The company has confirmed that Claude Code usage roughly doubled between January and April 2026 (per the May 2026 rate-limit expansion announcement) but has never published an absolute weekly user count. A growth rate disclosure without an anchor number is analytically unverifiable — it signals direction without enabling benchmarking. Cursor reports annual recurring revenue ($2 billion ARR reportedly reached by February 2026) rather than users. GitHub Copilot user data surfaces through Microsoft earnings calls, which follow quarterly financial reporting cycles.

Most coverage placing Codex at 4M next to GitHub Copilot at 20M fails to flag that these are different metric types measured over different windows. Codex's 4M is weekly active users — a relatively strict engagement filter. Copilot's 20M is all-time cumulative users, disclosed as of July 2025 and not updated since. A developer who tried Copilot twice in 2022 counts toward the 20M; a developer who uses Codex every day counts toward the 4M. One filters for recency and engagement; the other counts anyone who ever touched the product.

Similarly, comparing Codex 4M WAU to Cursor's 1M+ DAU (December 2025, via aggregator) conflates weekly with daily — and a daily active user metric typically implies higher engagement intensity per user, not lower penetration. The Cursor DAU figure is also aggregator-sourced rather than vendor-disclosed, adding another layer of uncertainty.

The honest read of the available data: Codex has grown dramatically from a small base over five months. Whether it has surpassed Copilot, Cursor, or Claude Code in any meaningful engagement-normalized sense is genuinely unknowable from current public disclosures. The only defensible comparison is the one OpenAI controls: the trajectory of its own WAU milestones over time.

For a deeper comparison of how all three major coding-agent platforms have tracked across the first half of 2026, including the caveats on every user-count claim, see our H1 2026 retrospective on Cursor, Claude Code, and Codex data.

When Codex hit 3 million WAU on April 8, 2026, Sam Altman announced that OpenAI would reset Codex usage limits — and pledged to repeat that action at every additional million-user milestone up to 10 million. The announcement coincided with the launch of the ChatGPT Pro $100/month tier, which offered 5x the Codex usage allowance of the Plus tier (temporarily boosted to 10x through May 31, 2026).

The pledge is unusual for a consumer software product. Usage-limit resets are typically one-off celebrations; committing to repeat them at defined milestones creates an implicit forward-curve forecast. If taken at face value, Altman was signaling a roadmap: Codex would reach at least 10M WAU, and the journey would be marked by a series of limit-expansion events designed to accelerate adoption at each new threshold.

At 4M WAU today, the Altman pledge implies 2.5x growth remains to reach the stated ceiling. Based on the January-to-April trajectory (roughly 3.4M added in ~16 weeks), the implied timeline to 10M under similar growth rates would be mid-to-late 2026 — though growth typically decelerates as the addressable market of early adopters is absorbed.

OpenAI's Q1 2026 revenue reached approximately $5.7 billion, according to The Information's reporting. Enterprise channels accounted for more than 40% of that total, with enterprise projected to reach parity with consumer revenue by the end of 2026. Codex is cited by The Information as one of several drivers of enterprise expansion — alongside ChatGPT Business and Enterprise API usage — but should not be read as the sole engine of that growth. Consumer ChatGPT, API revenue, and other surfaces all contribute to the $5.7B figure.

The enterprise framing matters for understanding the 4M WAU trajectory. Enterprise deployments typically add WAU in large, discrete chunks rather than the gradual consumer adoption curve. When a firm like Cisco or Accenture deploys Codex across its engineering organization, thousands of weekly active users can appear in the statistics simultaneously. This lumpy adoption pattern likely explains why the 3M-to-4M sprint (two weeks) was faster than the 2M-to-3M segment (three weeks) — a large enterprise signing rather than an organic consumer acceleration.

The ChatGPT Pro $100 tier, launched specifically on April 9 as Codex usage grew, also reflects enterprise orientation. At $100/month per seat, the tier targets professional developers and teams rather than individual hobbyists — and the 5x-Plus Codex allowance was the headline feature justifying the price. Per TechCrunch, this tier was Codex-led. For teams evaluating whether Codex belongs in your stack, an AI transformation engagement can structure the build vs buy analysis against your existing toolchain.

Today's partnership with Dell Technologies is the clearest signal yet that OpenAI views the current 4M WAU ceiling as a constraint of deployment model, not addressable market. The announcement describes Codex connecting with the Dell AI Data Platform and Dell AI Factory for hybrid and on-premises deployments — meaning enterprise customers whose data governance requirements preclude cloud-hosted inference can now run Codex inside their own infrastructure.

Dell's own framing is explicit about the motivation: “The Dell AI Factory with OpenAI Codex will allow enterprises to deploy AI where enterprise data already lives, within their premises, giving customers a practical, secure path to deploying AI agents at scale.” The key phrase is “where enterprise data already lives.” A meaningful segment of potential enterprise Codex adopters — regulated financial services, defence contractors, government, healthcare, and others operating under strict data residency requirements — has been blocked from cloud deployment entirely. The Dell partnership removes that blocker.

The strategic read: if OpenAI expected another 2-3x of cloud-native WAU growth from the current base, the on-prem partnership would be less urgent. The timing suggests OpenAI believes the easy SaaS growth has been captured and the incremental opportunity lies in harder-to-reach enterprise segments that require infrastructure flexibility. That is a growth-maturity transition, not a growth problem — but it does reframe what “the next million users” requires operationally.

For a deeper analysis of the Dell deal itself, see our dedicated post: OpenAI and Dell Codex on-premises enterprise deployment.

The Altman 10M pledge implies a 2.5x increase from the current base. Extrapolating the January-to-April growth rate of approximately 3.4M new WAU across 16 weeks, the 10M ceiling would be reachable around Q3 or Q4 2026 under continued similar growth. In practice, growth rates compress as the early-adopter pool is absorbed — the more realistic scenario is that 10M represents a 12-to-18-month horizon from the April 21 disclosure, conditional on the on-prem channel opening meaningfully.

The margin implication matters more than the milestone itself. Codex at 4M WAU runs primarily on gpt-5-codex, gpt-5.1-codex, gpt-5.2-codex, and gpt-5.3-codex model variants — each of which carries significant inference cost at scale. As WAU doubles, the cost of delivering free-tier and Plus-tier Codex usage grows in proportion, unless the underlying inference cost per token falls (driven by hardware improvements or model compression) or OpenAI successfully migrates usage toward paid tiers. The April 9 Pro $100 tier and the existing Pro $200 tier are the current monetization instruments for this transition.

The enterprise channel changes the margin math more favorably. Enterprise contracts price Codex at substantially higher per-seat rates than consumer tiers, and Dell on-prem deployments shift some compute cost entirely onto customer hardware — improving OpenAI's unit economics on those users. If the next million WAU are predominantly enterprise or on-prem, OpenAI can grow the user base while improving, not compressing, its margin structure per incremental user.

For teams modeling where to place AI coding investment across 2026 and into 2027, the 2026 developer-survey adoption data and the Codex desktop computer-use and plugins guide provide the practical deployment context behind the WAU headline.