I’ve used Claude Code since mid-2025 and have now spent more than 800 hours working with it. That experience cannot tell me how many developers use the tool or how much more productive they are with it. Those questions need the right measurements.
So I separate downloads, surveys, revenue announcements, benchmarks, and analyzed sessions. Each number answers a different question.
- In JetBrains' 2026 survey, 39% of more than 15,000 professional developers said they use Claude Code at work. That is not market share.
- Anthropic reported Claude Code revenue run rate above $2.5 billion on February 12, 2026. That is an annualized pace, not actual annual revenue.
- The Claude Code npm package reached 562.7 million downloads through October 6, 2026. Downloads are not unique users.
1. What Is Claude Code?
Claude Code is Anthropic’s coding agent. It can read a project, edit files, run commands, and manage changes with Git. Anthropic offers it in the terminal, IDEs, a desktop app, and a browser. The official documentation describes its available surfaces and features.
The main public milestones show the difference between product preview and general availability:
The preview launched alongside Claude 3.7 Sonnet. General availability followed in May. Those dates appear in Anthropic’s February 24, 2025 announcement and its May 22, 2025 Claude 4 announcement. The version number comes from the GitHub releases.
2. GitHub Stars and npm Downloads
The GitHub repository had 149,665 stars according to the GitHub API on October 7, 2026. Stars show interest in the repository, but they do not tell us how many people actively use Claude Code.
- 149,665 GitHub stars on October 7, 2026
- 562,748,153 npm downloads from February 24, 2025, to October 6, 2026
- 52,946,380 npm downloads in the 30-day period from September 7 to October 6, 2026
- Version 2.1.292, released October 6, 2026
The 30-day figure comes from the npm download API. The cumulative total in the chart adds two npm API date ranges because a single API request covers no more than 18 months. According to npm, the count tracks HTTP downloads of package archives. It includes automated build servers, mirrors, and bots, so it is not a count of unique active users.
Sources: npm API, February 2025 to August 2026, npm API, August to October 2026. February 2025 covers only the period from February 24. The October 2026 total is partial through October 6.
3. User Numbers and Adoption
Anthropic has not published an absolute count of active Claude Code users. The 2026 Developer Ecosystem Survey from JetBrains offers a current comparison of reported use at work. Its tenth edition surveyed more than 15,000 professional developers worldwide. JetBrains describes the survey as globally representative and weights it by factors including region and employment status. Across all AI coding agents, 90% of respondents reported using one at least weekly and 68% daily.
Self-reported use at work. January and May to July are separate survey waves.
In the summer wave, 39% of respondents reported using Claude Code at work. Thirty-one percent called it their main AI coding tool. Reported use in the United States was 47%. The 18% figure from January comes from a separate survey wave. The data do not track the same people over time. Source: the JetBrains Developer Ecosystem Survey 2026.
4. Market Share or Survey Results?
Survey figures are not market shares. They measure what a specific group of respondents reports using. They do not count paid contracts, revenue, or all active accounts worldwide.
The surveys use different samples and questions. Their results are not directly comparable.
The Pragmatic Engineer survey ran from January 27 to February 17, 2026, among newsletter readers. A total of 507 people, or 55% of respondents, regularly used AI agents. Within that subgroup, 71% named Claude Code. This was a self-selected newsletter sample, not a global user count. The Pragmatic Engineer survey report also notes that respondents could name several tools.
5. Popularity and Satisfaction
In the same Pragmatic Engineer survey, 46% named Claude Code as a tool they especially enjoy using. Cursor came in at 19%, and GitHub Copilot at 9%. The question asked which AI tools respondents loved using most and why.
This is a signal about preferences among the surveyed newsletter readers. It is not a global satisfaction measure, a recommendation score, or market share. Because respondents could name several tools, the percentages are not exclusive and do not need to add up to 100%.
6. Revenue and Business Figures
Anthropic publishes some revenue figures as run rates. Bloomberg describes this metric as projecting a shorter period across a full year. It is not the same as revenue actually booked during a calendar year, and it does not by itself establish contractually recurring revenue in the sense of ARR.
For Claude Code alone, the latest separately stated run rate I found in a public company source was more than $2.5 billion on February 12, 2026. Anthropic said it had more than doubled since the start of the year. It also said weekly active Claude Code users had doubled since January 1, but did not provide an absolute user count. Those figures appear in Anthropic’s February 12 announcement. Anthropic had reported a Claude Code run rate above $500 million in August 2025, then above $1 billion in November, six months after general availability. The $500 million milestone was announced on September 2; the $1 billion milestone followed in November.
The company figures cover Anthropic overall, not Claude Code alone. Anthropic reported a $14 billion run rate on February 12 and more than $47 billion earlier in May in its May 28 announcement. Bloomberg reported on August 18 that it was above $65 billion at the end of July, citing anonymous sources. On September 18, Axios cited a New York Times report that Anthropic was pacing toward more than $100 billion in annualized revenue for 2026. The later figures are media reports about private company data or expectations, not audited annual accounts. They cannot be attributed to Claude Code revenue.
7. Coding Benchmarks
Anthropic publishes Terminal-Bench 4.0 results for its coding models. These are tests under defined conditions. They do not predict how reliably Claude Code will handle every individual task.
Provider results. The effort settings are not identical, so these figures are not a direct ranking.
Opus 5.5 was released on September 22, 2026, and Sonnet 5.5 on September 28. Anthropic specifies Xhigh effort for the Opus 5.5 result and reports a 2.6 percentage point standard error on Terminal-Bench 4.0. The Sonnet test used Claude Code in “--bare” mode at maximum effort, across 66 tasks and five trials per task, without internet access. Anthropic reports a 2.5 percentage point standard error; safeguards and server-side fallback models affected a small share of trials. Anthropic’s Opus release report and Sonnet system card describe the respective test setups. These are model benchmarks, not measures of productivity among Claude Code users, and the different setups do not support a direct ranking.
8. What We Know About Successful Work and Productivity
An Anthropic study analyzed about 400,000 interactive sessions from approximately 235,000 people between October 2025 and April 2026. It covered Claude Code in the terminal, Claude.ai, and the desktop app. It did not cover usage through third-party IDEs, the SDK, or headless mode.
Share of sessions, not share of people. Anthropic study published June 16, 2026, covering October 2025 to April 2026.
A session counted as a verified success only if a classifier judged its goal to have been met and there was a verifiable signal, such as a relevant commit, pull request, passing tests, or explicit user confirmation. Anthropic excluded sessions without a clear goal from this success analysis. They represented about 7.7% of the full sample.
Anthropic also says the study could not observe whether results were later used or discarded. It does not measure time saved or a causal productivity gain. These rates describe classified sessions with code changes, not users. Source: the Anthropic study of Claude Code in practice.
Acceptance rates are not the same as productivity. Anthropic’s Claude Code Analytics API gives organizations daily metrics, including accepted or rejected proposals for file-editing tools, sessions, lines changed, commits, and pull requests. An accepted edit alone does not show whether it was correct, shipped, or saved time. The API documentation does not publish a globally representative acceptance rate across all Claude Code users.
9. Pricing and Costs
Claude Code is not included in Claude’s free plan. Individual access starts with Pro. Max offers higher usage limits. Anthropic’s US price list distinguishes monthly and annual billing.
US list prices, checked October 7, 2026. Taxes may apply. Usage limits apply.
Annual Pro billing works out to about $17 per month, but the $200 annual fee is billed up front. Team and Enterprise plans also include Claude Code, with different pricing and usage terms. Anthropic’s pricing page says taxes and usage limits may apply.
API pricing depends on the model and number of input and output tokens. Anthropic lists these standard rates per million tokens for the two current coding models:
Prompt caching and Opus Fast mode have separate prices. The standard rate alone therefore cannot determine the cost of a task. Current details are in Anthropic’s API pricing documentation.
10. Technical Specifications
Some commonly quoted product details depend on the version, model, or plan. Those differences are part of the figures.
The context values come from the individual model overviews for Claude Opus 5.5 and Claude Sonnet 5.5. They are not a guarantee for every request. For installation, supported surfaces, and operating systems, see Anthropic’s Claude Code documentation and plan overview. The GitHub repository’s license file refers to Anthropic’s commercial terms.
11. What These Numbers Do and Do Not Tell Us
GitHub stars show interest in a repository. npm downloads count package requests. Surveys show self-reported use in a particular sample. Run rates annualize a current revenue pace. Benchmarks score models under test conditions. Acceptance rates count proposals that users approve. None of these figures alone shows how much time an individual saves with Claude Code or how many changes make it into production.
The clearest way to read a statistic is to keep its measure, period, sample, and source visible. That is why the public figures for Claude Code users, acceptance rates, and productivity effects do not appear here as one ranking or percentage.
If you want to try Claude Code, our Claude Code guide has a place to start. For general figures about Claude and Anthropic, see Claude Statistics 2026.
Frequently Asked Questions About Claude Code
Under US pricing, Pro costs $20 with monthly billing or $200 paid up front for a year (effectively $17 per month). Max costs $100 per month for the 5x tier or $200 per month for the 20x tier. API usage is billed separately by model and token count.
For a detailed breakdown of plan tiers and API costs, see the Claude Code pricing guide.
